JP4453046B2 - Vehicle behavior learning apparatus and vehicle behavior learning program - Google Patents

Description

  The present invention relates to a vehicle behavior learning device and a vehicle behavior learning program capable of learning a specific behavior of a vehicle at a specific vehicle position.

  When a specific driver, such as a private car, frequently drives the same route, a specific behavior may be frequently performed at a specific position of the vehicle. The behavior includes, for example, turning left or right when heading to a specific place such as home, work, or shop, opening / closing a window, turning on / off a light, kicking down in an auto-matching transmission. In recent years, a navigation system that performs route guidance is mounted on many vehicles, and it is desired to further improve the accuracy of the navigation system. Further, various studies have been made to expand various functions of the navigation system to applications other than route guidance. Japanese Patent Application Publication No. JP-A-2001-259542, which is cited below, describes an invention relating to a control device for a blind spot monitor in a vehicle equipped with a navigation system.

  This blind spot monitor control device stores, as activation information, data relating to the point at which the vehicle is input, which is input from the navigation system, when a manual switch for activating the blind spot monitor is operated. Then, when the vehicle is located at the point after searching and collating the activation information with respect to the input information from the navigation system, a blind spot monitor activation signal is transmitted. This navigation system manages road information using lines (links) connecting coordinate points such as intersections. When the point where the manual switch is operated is a road with a link number, the activation information is stored including the link number, coordinates, and the traveling direction of the vehicle. If the point is other than a road with a link number, activation information including coordinates and the traveling direction of the vehicle is stored. Further, the location where the vehicle is located is determined by using a GPS system and a hybrid system that estimates by autonomous navigation based on a vehicle speed signal and an angular velocity signal.

JP 2002-286458 A (see paragraphs 1 to 14, paragraphs 24 to 28, etc.)

  The blind spot monitor control device described in Patent Literature 1 obtains the position of the host vehicle using a hybrid system. However, as described in Japanese Patent Laid-Open No. 2004-228561, whether the positioning is performed by GPS positioning or autonomous navigation, the positioning value includes an error and deviates from the actual traveling position. Therefore, the most likely position is estimated by so-called map matching processing.

  By the way, not only the activation operation of the blind spot monitor but various specific behaviors in the vehicle are often performed at specific points. For example, a left-right turn from a large road to a narrow street, kick-down, etc. are performed when approaching a specific place such as a home, a workplace or a store. There are many narrow streets that diverge from large roads, and the intervals are often narrow. When the narrow street interval is smaller than the positioning value error, it is difficult to predict the behavior of the vehicle from the positioning result.

  The present invention was devised in view of the above problems, and is a vehicle behavior learning device and vehicle behavior learning capable of learning a high-frequency vehicle behavior at a specific position on a road with high positional accuracy. The purpose is to provide a program.

The characteristic configuration of the vehicle behavior learning apparatus according to the present invention for achieving the above object includes: own vehicle position information acquisition means for acquiring own vehicle position information indicating the current position of the own vehicle; and image information around the own vehicle. Image information acquisition means for acquiring, feature information storage means for storing feature information including position information and attribute information for a plurality of target features, and the feature information storage based on the vehicle position information A feature information acquisition unit that acquires the feature information around the vehicle from the unit, an image recognition unit that performs a recognition process of the target feature included in the image information based on the feature information, and the image A behavior detecting means for detecting the behavior of the host vehicle within a predetermined range from a position at which the image of the target feature has been successfully recognized by the recognition means; and a detection of the behavior of the host vehicle by the behavior detecting means based on the host vehicle position information. Behavior detection information indicating the result And a plurality of the same behavior stored in the detection result storage means when the host vehicle passes through the same place a plurality of times. Based on the behavior detection information, the behavior of the host vehicle that is repeatedly detected is extracted as the learning behavior, and the attribute information and the position information of the learning behavior are associated with the feature information of the target feature that has succeeded in the image recognition. Learning behavior extraction means for outputting as learning behavior information, and the learning behavior extraction means is communicably connected to a plurality of vehicles, and the behavior detection means is externally applied to the own vehicle from the outside as the behavior of the own vehicle. When detecting the movement of the own vehicle due to a factor, a plurality of vehicle behaviors stored in the detection result storage means of each vehicle when the plurality of vehicles pass through the same place. Based on the behavior detection information, certain behavior of the vehicle detected with a reproducible point extracted as the learned behavior.

According to this feature configuration, the learning behavior information includes the behavior detection information in which the behavior of the host vehicle is detected within a predetermined range from the position at which image recognition of the target feature is successful, and the information on the detection position in which the behavior is detected. including. That is, the target feature existing at a specific position and the behavior of the host vehicle are associated and output as learning behavior information. In addition, the learning behavior information includes the behavior of the own vehicle that is repeatedly detected at a location based on the target feature existing at the specific position, and thus indicates the behavior of the own vehicle that frequently appears at the specific position. It will be a thing. Compared with positioning based only on GPS positioning or autonomous navigation, high positioning accuracy can be realized by using a positioning method based on recognition of target features. Therefore, the learning behavior information of this feature configuration associated with the target feature also has high position accuracy. As a result, it is possible to provide a vehicle behavior learning device capable of learning the behavior of a vehicle with a high frequency at a specific position on the road with high positional accuracy. Further, by using the learning behavior information output from the learning behavior extracting means, it is possible to predict the behavior of the vehicle with reference to the target feature that has been image-recognized.
By the way, the behavior of the own vehicle due to an external factor is a behavior that occurs in common in a plurality of other vehicles even if there is a difference in degree. In other words, it occurs due to passive causes such as road conditions and weather, so even if the same behavior is not repeatedly detected in the same vehicle, it can be used as learned behavior information if the same behavior is reproduced in multiple vehicles. it can. According to this feature, since the learning behavior extracting means is communicably connected to a plurality of vehicles, even if the behavior detected in the own vehicle is one time, it is matched with the behavior detected in other vehicles. Thus, a reproducible behavior can be obtained. As a result, it is possible to learn a wide range of vehicle behaviors.

Here, wherein the behavior of the host vehicle detected by the behavior detecting means includes operation of at least the vehicle, the reception of the driver's operation in each part of the vehicle is also preferable in I free.
In the present application, the operation of the host vehicle includes the operation of each part of the host vehicle and the entire vehicle caused by the operation of the driver, or the operation of each part of the host vehicle and the entire vehicle caused by an external factor applied to the host vehicle from the outside. Etc. are included.

  There are two types of behavior of the host vehicle: one that occurs in response to a driver's own operation, and one that occurs due to passive causes such as road conditions and weather. For example, the driver's spontaneous operation includes a steering operation, a shift operation, an accelerator or brake operation, an air conditioner operation, a window opening / closing operation, a navigation device or an audio operation. As the behavior of the host vehicle caused by such a driver's spontaneous operation, for example, a change in the traveling direction of the host vehicle or a change in lateral acceleration caused by a steering operation, a shift operation or an accelerator operation, etc. There are changes in the gear position (kick down, shift down, etc.), changes in the acceleration in the traveling direction caused by the operation of the brake and accelerator, and the like. On the other hand, the behavior of the host vehicle caused by passive causes includes, for example, changes in lateral acceleration caused by running on a curve, vibrations / impacts caused by stepping, changes in acceleration in the traveling direction when running on a slope, etc. is there. The behavior detecting means according to the present invention detects at least one of the autonomous operation of the driver and the operation of the own vehicle due to the driver's operation or an external factor as the behavior of the own vehicle. The behavior of the vehicle can be detected.

  In addition, it is preferable that the vehicle behavior learning device according to the present invention includes behavior prediction means for predicting the behavior associated with the target feature based on the learned behavior information.

  According to this configuration, as described above, the learning behavior information indicating the behavior of the vehicle with a high frequency at a specific position on the road learned with high position accuracy is used to predict the behavior. Can do. That is, it is possible to accurately predict the behavior of the host vehicle on the basis of the image-recognized target feature.

  When the vehicle behavior learning device according to the present invention includes the behavior predicting unit, the behavior predicting unit outputs the prediction result of the behavior to a navigation calculating unit that performs arithmetic processing for outputting guidance information of the host vehicle. Is preferably output.

  According to this configuration, based on the prediction result of the behavior prediction unit, the navigation calculation unit can output appropriate guidance information. For example, it is possible to predict the behavior of the host vehicle such as a right turn or a left turn, and to accurately display the host vehicle position and route guidance.

  When the vehicle behavior learning device according to the present invention includes the behavior prediction unit, it is preferable that the behavior prediction unit outputs the behavior prediction result to a control device in the host vehicle that reproduces the operation of the driver. It is.

  According to this structure, the said control apparatus can reproduce a driver | operator's operation based on the prediction result of a behavior prediction means. That is, it is possible to assist the driver in active operation, or to perform the same operation prior to the active operation. As a result, the behavior of the vehicle can be reliably reproduced even if the active operation by the driver is delayed or forgotten.

  When the vehicle behavior learning device according to the present invention includes the behavior prediction unit, the behavior prediction unit preferably outputs the behavior prediction result to a control device that optimizes the operation of the host vehicle.

  According to this structure, the said control apparatus can optimize the behavior of the own vehicle based on the prediction result of a behavior prediction means. For example, a shift control device, an engine control device, a motor control device for a hybrid vehicle, a control device for an active suspension, and the like can optimize the behavior of the host vehicle based on the prediction result of the behavior prediction means. As a result, it is possible to improve comfort while riding and improve fuel efficiency.

  Here, it is preferable that the target feature is a road marking provided on a road surface.

  The dimensions of road markings are regulated by law. Therefore, it is possible to reduce the load of recognition processing by the image recognition means and to perform accurate position determination. That is, the position information of the host vehicle can be obtained with high accuracy. As a result, it is possible to provide a vehicle behavior learning device capable of learning the behavior of the vehicle with high positional accuracy.

Further, in the above feature information storing means, wherein the feature information, previously stored initial feature information,及beauty before Symbol image recognition unit by Ri the feature information and the feature that place independently in It is preferable that one or both of the learned feature information learned and stored based on the result of the image recognition process is stored.

  According to this configuration, when the initial feature information is stored in the feature information storage means, it is possible to perform vehicle behavior learning using the initial feature information, and the learned feature information If stored, it is possible to perform vehicle behavior learning using the learned feature information. When both the initial feature information and the learned feature information are stored, it is possible to perform vehicle behavior learning using one or both of them according to the situation. In addition, in the configuration in which learning feature information is stored, it is possible to sequentially prepare feature information for roads on which the vehicle has traveled, even on roads for which initial feature information is not maintained. It becomes. Therefore, it is possible to expand roads and areas where vehicle behavior learning can be performed according to the travel route of the host vehicle, and it is also possible to reduce labor and cost for maintaining initial feature information.

  The information on the detection position where the behavior is detected includes coordinate information indicating the detection position of the behavior, and the position of the target feature that has succeeded in the image recognition to the detection position of the behavior. A configuration including one or both of the distance information is preferable.

  According to this configuration, the information on the detection position of the behavior, the information indicating the detection position alone, and the distance between the feature and the behavior indicating the relationship between the detection position and the target feature that has succeeded in the image recognition. It is possible to appropriately express using one or both of the information represented by. Further, in the case of the configuration including the distance information, using the distance information, the distance information from the position of the target feature that has succeeded in the image recognition to the position of the learning behavior is used as the position information of the learning behavior. The learning behavior information including it can be easily generated and output by the learning behavior extraction means.

  Further, the position information of the learning behavior related to the learning behavior information is one of coordinate information representing the position of the learning behavior and distance information from the position of the target feature that has succeeded in the image recognition to the position of the learning behavior. Or it is suitable if it is set as the structure containing both.

  According to this configuration, the position information of the learning behavior, the information indicating the position of the learning behavior alone, and the feature-learning indicating the relationship between the position of the learning behavior and the target feature that has succeeded in the image recognition It is possible to appropriately represent one or both of the information represented by the distance between the behaviors. Further, in the case of the configuration including the distance information, when the image recognition of the target feature indicated in the learning behavior information is successful, it is more accurate based on the position where the image recognition of the target feature is successful. It is possible to predict the occurrence of learning behavior.

  In addition, the characteristic configuration of the navigation device according to the present invention is output by the vehicle behavior learning device having any of the above-described characteristic configurations, the map information storage means in which the map information is stored, and the vehicle behavior learning device. And a plurality of application programs that operate with reference to one or both of the learning behavior information and the map information, and guide information output means that operates according to the application program and outputs guide information.

  According to this feature configuration, the application program can operate using learning behavior information learned with high accuracy. Therefore, the guide information can be output with higher accuracy. For example, if bending behavior on frequently used narrow streets such as home is learned as a behavior, map matching to narrow streets can be performed accurately without accidental map matching to large streets near narrow streets can do.

Further, the vehicle behavior learning program according to the present invention is characterized by a vehicle position information acquisition step for acquiring own vehicle position information indicating the current position of the own vehicle, and image information acquisition for acquiring image information around the own vehicle. The feature information around the own vehicle is acquired based on the own vehicle position information from the feature information storage means storing the feature information including position information and attribute information about the plurality of target features. A feature information acquisition step, an image recognition step for recognizing a target feature included in the image information based on the feature information, and a position at which image recognition of the target feature by the image recognition means is successful. The behavior detection step for detecting the behavior of the host vehicle within a predetermined range from the host vehicle and the behavior detection information representing the detection result of the behavior of the host vehicle by the behavior detection means are detected based on the host vehicle position information. Of the detected position A detection result storing step for storing in the detection result storage means in association with the information, and a plurality of behavior detections for the behavior of the same own vehicle stored in the detection result storage means when the own vehicle passes the same place a plurality of times Based on the information, the behavior of the subject vehicle that is repeatedly detected is extracted as the learning behavior, and the learning behavior detection information that associates the attribute information and the position information of the learning behavior with the feature information of the target feature that has succeeded in the image recognition. A learning behavior extraction step that is output as a computer, and in the learning behavior extraction step, the behavior detection step detects the behavior of the host vehicle due to an external factor that is externally applied to the host vehicle as the behavior of the host vehicle. , Communicating with a plurality of vehicles, and the plurality of vehicles passing through the same place and stored in the detection result storage means of each vehicle, Based on the number of behavior detection information, certain behavior of the vehicle detected with a reproducible point extracted as the learned behavior.

According to this feature configuration, the learning behavior information includes the behavior detection information in which the behavior of the host vehicle is detected within a predetermined range from the position at which image recognition of the target feature is successful, and the information on the detection position in which the behavior is detected. including. That is, the target feature existing at a specific position and the behavior of the host vehicle are associated and output as learning behavior information. In addition, the learning behavior information includes the behavior of the own vehicle that is repeatedly detected at a location based on the target feature existing at the specific position, and thus indicates the behavior of the own vehicle that frequently appears at the specific position. It will be a thing. Compared with positioning by only GPS positioning or autonomous navigation, higher positioning accuracy can be realized by using a positioning method by recognizing a target feature. Therefore, the learning behavior information of this feature configuration associated with the target feature also has high position accuracy. As a result, it is possible to provide a vehicle behavior learning program capable of learning the behavior of a vehicle with a high frequency at a specific position on the road with high positional accuracy. Moreover, since it communicates with several vehicles in a learning behavior extraction process, even if the behavior detected in the own vehicle is 1 time, it is set as the behavior which has reproducibility by combining with the behavior detected in other vehicles. be able to. As a result, it is possible to learn a wide range of vehicle behaviors.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram schematically showing a configuration example of a navigation device 1 including a vehicle behavior learning device 2 according to the present embodiment. FIG. 2 is an explanatory diagram showing a configuration example of information stored in the map database 13 and the feature database 15 shown in FIG. The map database 13 corresponds to the map information storage means of the present invention, and the feature database 15 corresponds to the feature information storage means of the present invention. As shown in FIG. 1, the navigation device 1 includes functional units of a vehicle behavior learning device 2, a map database 13, an application program 16, and a guidance information output unit (guidance information output means) 28. . The vehicle behavior learning device 2 includes a host vehicle position information acquisition unit 3, an image information acquisition unit 4, a feature information acquisition unit 5, an image recognition unit 6, a behavior detection unit 7, a detection result storage unit 8, And a learning behavior extracting unit 9. These functional units are constructed by one or both of hardware and software (program) using an arithmetic processing unit such as a microprocessor or a DSP (digital signal processor) as a core member. That is, each of these functional units executes processing on various input data by cooperation of hardware and software. Details of each functional unit will be described later.

[Map database and feature database]
The map database 13 is a database in which a plurality of map information M divided for each predetermined area and a plurality of feature information F associated with the map information M are stored. FIG. 2 shows a configuration example of the map information M stored in the map database 13 and the feature information F stored in the feature database 15. As shown in the figure, the map database 13 stores a road network layer m1, a road shape layer m2, and a feature layer m3. In the present embodiment, the map information M of the map database 13 is configured by the information stored in each of these layers m1 to m3, and the features of the feature database 15 are configured by the information stored in the feature layer m3 therein. Information F is configured.

  The road network layer m1 is a layer indicating connection information between roads. Specifically, information on a large number of nodes n having position information on a map expressed by coordinates such as latitude and longitude, and information on a large number of links k constituting the road by connecting the two nodes n. It is configured. Each link k includes information such as the type of road (type of highway, toll road, national road, prefectural road, etc.) and link length as link information. The road shape layer m2 is a layer that is stored in association with the road network layer m1 and indicates the shape of the road. Specifically, information on a number of road shape complementary points s having position information on a map arranged between two nodes n (on a link k) and expressed by coordinates such as latitude and longitude, and road width w information and the like.

  The feature layer m3 is configured in association with the road network layer m1 and the road shape layer m2, and is a layer in which information on various features provided on the road and around the road, that is, the feature information F is stored. is there. The feature in which the feature information F is stored in the feature layer m3 includes a road marking (paint marking) provided on the road surface. Examples of features related to such road markings include, for example, lane markings that divide lanes along the road (including various lane markings such as solid lines, broken lines, and double lines), and progressions that specify the traveling direction of each lane. Includes direction-specific traffic markings, pedestrian crossings, stop lines, speed indications, zebra zones, and more. In addition to such road markings, the features in which the feature information F is stored can include various features such as traffic lights, signs, overpasses, and tunnels.

  The feature information F includes position information and attribute information of each feature as its contents. Here, the position information includes information on the position (coordinates such as latitude and longitude) of the representative point of each feature associated with the link k or the node n on the map and the direction of each feature. In this example, the representative point is set at the center in the length direction and the width direction of each feature. The attribute information includes form information representing the form of each feature, type information representing the type of each feature, and the like. Here, the form information includes information such as the shape, size, and color of each feature. The type information specifically includes information indicating road marking types such as “division line (including line types such as solid line, broken line, double line)”, “traveling direction indication according to traveling direction”, “pedestrian crossing”, etc. It is.

[Vehicle position calculation unit]
The own vehicle position information acquisition unit 3 functions as own vehicle position information acquisition means for acquiring own vehicle position information L indicating the own vehicle position, that is, the current position of the own vehicle. In this example, the vehicle position information acquisition unit 3 is connected to the GPS receiver 23, the direction sensor 24, and the distance sensor 25.
The GPS receiver 23 is a device that receives GPS signals from GPS (Global Positioning System) satellites. This GPS signal is normally received every second and output to the vehicle position information acquisition unit 3. The own vehicle position information acquisition unit 3 analyzes the received signal from the GPS satellite, and acquires information such as the current position of the own vehicle (coordinates such as latitude and longitude), traveling direction, and moving speed.
The direction sensor 24 detects the traveling direction of the host vehicle or a change in the traveling direction and outputs the detection result to the host vehicle position information acquisition unit 3. The azimuth sensor 24 includes, for example, a gyro sensor, a geomagnetic sensor, an optical rotation sensor attached to the rotating part of the handle, a rotary resistance volume, an angle sensor attached to the wheel part, and the like.
The distance sensor 25 detects the vehicle speed and travel distance of the host vehicle, and outputs information on the vehicle speed and travel distance as a detection result to the host vehicle position information acquisition unit 3. The distance sensor 25 integrates, for example, a vehicle speed pulse sensor that outputs a pulse signal every time a drive shaft or wheel of the vehicle rotates a certain amount, a yaw / G sensor that detects the acceleration of the host vehicle, and the detected acceleration. It is composed of a circuit or the like.

  Based on the outputs from the GPS receiver 23, the azimuth sensor 24, and the distance sensor 25, the own vehicle position information acquisition unit 3 performs a calculation for specifying the own vehicle position by a known method. The vehicle position information acquisition unit 3 adjusts the vehicle position on the road indicated by the map information M by performing known map matching based on the map information M around the vehicle position acquired from the map database 13. To do.

  The information on the vehicle position acquired in this way may include an error caused by the detection accuracy of each of the sensors 23 to 25. In addition, the information on the vehicle position acquired in this way may not be able to accurately identify the lane in which the host vehicle is in progress when the road on which the host vehicle is in progress has multiple lanes. is there. Therefore, in the present embodiment, lane specifying information J specifying the lane in which the host vehicle is present on the road where the host vehicle is traveling is supplied to the host vehicle position information acquiring unit 3 from a lane specifying unit 17 described later. The own vehicle position information acquisition unit 3 uses the calculation result for specifying the own vehicle position and the lane identification information J supplied from the lane identification unit 17 to indicate the current position of the own vehicle represented by coordinates such as latitude and longitude. , Information on the traveling direction of the host vehicle, and host vehicle position information L including lane identification information J regarding the lane in which the host vehicle is traveling. The vehicle position information L is output to the vehicle position information correction unit 11, the feature information acquisition unit 5, the recognition condition setting unit 14, and the navigation calculation unit 12.

[Image information acquisition unit]
The image information acquisition unit 4 functions as an image information acquisition unit that acquires image information G around the vehicle position captured by the imaging device 21. Here, the imaging device 21 is a camera or the like provided with an imaging element, and is provided at a position where at least a road surface of a road around the own vehicle (own vehicle position) can be imaged. As such an imaging device 21, for example, a back camera or the like can be used. The image information acquisition unit 4 captures imaging information captured by the imaging device 21 at predetermined time intervals via a frame memory (not shown). At this time, the time interval for capturing the image information G can be set to about 10 to 50 ms, for example. Thereby, the image information acquisition unit 4 continuously acquires image information G of a plurality of frames captured by the imaging device 21. The image information G acquired here is output to the image recognition unit 6.

[Feature Information Acquisition Department]
The feature information acquisition unit 5 functions as a feature information acquisition unit that extracts the feature information F of the target feature ft to be subjected to image recognition processing from the feature database 15. In this example, the feature information acquisition unit 5 uses the ground information for one target feature ft set by the recognition condition setting unit 14 to be described later in order to use it for the process of correcting the position in the traveling direction of the own position information L. Object information F is acquired from the feature database 15. The feature information F acquired here includes position information and form information of the target feature ft as described above. The feature information F of the target feature ft extracted by the feature information acquisition unit 5 includes an image recognition unit 6, a vehicle position information correction unit 11, a detection result storage unit 9, a behavior prediction unit 10, and a recognition condition setting unit 14. Is output.

  In addition, the feature information acquisition unit 5 uses the lane identification unit 17 to acquire the lane identification information J, and based on the own position information L, the area around the own vehicle position on the road on which the own vehicle is traveling Line feature information F is extracted from the feature database 15. The lane marking feature information F extracted by the feature information acquisition unit 5 is output to the image recognition unit 6 and the lane identification unit 17.

[Recognition condition setting section]
The recognition condition setting unit 14 functions as a recognition condition setting unit that sets a recognition condition for image recognition processing in the image recognition unit.
The recognition condition setting unit 14 sets a target feature ft to be subjected to image recognition processing. The target feature ft is selected from among the features whose feature information F is stored in the feature database 15, from one or more features in the vicinity of the own position that falls within the imaging range of the imaging device 21. In this example, the recognition condition setting unit 14 is proceeding based on the vehicle position information L acquired by the vehicle position information acquisition unit 3 and the feature information F stored in the feature database 15. One feature existing at the closest position in the traveling direction in the lane is set as the target feature ft. In the present example, the range in which the target feature ft is searched is defined within a predetermined distance in the traveling direction of the host vehicle. Therefore, the target feature ft is not set when there is no feature within a predetermined distance in the traveling direction of the host vehicle.

  In addition, the recognition condition setting unit 14 sets an image recognition range that is a range in which image recognition processing is performed on each target feature ft with respect to the image information G. In the present embodiment, the image recognition range is a range defined in the traveling direction of the host vehicle (host vehicle position). This range is set according to the length of the target feature ft in its traveling direction and is obtained from the form information included in the feature information F. For example, a longer range is set when the target feature ft is longer in the traveling direction of the host vehicle, such as speed display, than when the target feature ft is a stop line. Information on the set image recognition range is output to the image recognition unit 6. Then, the image recognition unit 6 performs image recognition processing of the target feature ft on the image information G within the set image recognition range.

(Image recognition unit)
The image recognition unit 6 functions as an image recognition unit that performs image recognition processing on the image information G acquired by the image information acquisition unit 4. In the present embodiment, the image recognition unit 6 performs image recognition processing on the image information G within the range defined by the image recognition range set by the recognition condition setting unit 14. At this time, the image recognition unit 6 performs image recognition processing of the target feature ft using the feature information F of the target feature ft extracted by the feature information acquisition unit 5. Specifically, the image recognition unit 6 extracts image information G in the image recognition range from the image information G acquired by the image information acquisition unit 4. The information of the imaging region in each image information G is obtained based on the positional relationship between the own vehicle position and the imaging region calculated in advance based on the calibration information of the imaging device 21 acquired in advance and the own location information L. Can do. Here, the calibration information of the imaging device 21 is information (translation / rotation / camera internal parameters) based on the mounting position, the mounting angle, and the angle of view of the imaging device 21 to the host vehicle.

  The image recognizing unit 6 extracts the image information G in the image recognition range based on the information of the imaging area of each image information G obtained in this way. Specifically, the image recognition unit 6 performs binarization processing, edge detection processing, and the like on the extracted image information G, and outline information of the feature (road marking) included in the image information G To extract. After that, the image recognition unit 6 compares the extracted contour information of the feature with the form information included in the feature information F of the target feature ft acquired by the feature information acquisition unit 5, and they match. It is determined whether or not to do. If the contour information of the feature matches the form information included in the feature information F of the target feature ft, it is determined that the image recognition of the target feature ft has succeeded, and the image recognition result is determined as the vehicle position. The information is output to the information correction unit 11. If the image recognition of the target feature ft fails, the image recognition result is not output to the own vehicle position information correction unit 11, and therefore the own position information L is also corrected by the own vehicle position information correction unit 11. I will not.

  Further, the image recognition unit 6 is configured to acquire the lane identification information J in the lane identification unit 17 so that the lane markings around the vehicle position on the road on which the vehicle extracted by the feature information acquisition unit 5 is traveling are displayed. Using the object information F, image recognition of the lane markings around the host vehicle is performed. Specifically, the image recognition unit 6 performs binarization processing, edge detection processing, and the like on the image information G acquired by the image information acquisition unit 4, and the features ( (Road marking) contour information is extracted. Thereafter, the image recognition unit 6 determines the position of the lane line and the type of the lane line around the own vehicle based on the extracted contour information of the feature and the form information included in the lane line feature information F. Recognize. Then, the image recognition unit 6 outputs the image recognition result of the lane markings to the lane specifying unit 17.

[Vehicle position information correction unit]
The own vehicle position information correction unit 11 is based on the result of the image recognition process by the image recognition unit 6 and the position information of the target feature ft included in the feature information F acquired by the feature information acquisition unit 5. It functions as own position information correction means for correcting the position information L. In the present embodiment, the vehicle position information correction unit 11 corrects the vehicle position information L along the traveling direction of the vehicle. Specifically, the own vehicle position information correction unit 6 first acquires the image information G including the image of the target feature ft based on the image recognition result by the image recognition unit 6 and the calibration information of the imaging device 21. The positional relationship between the host vehicle and the target feature ft at the time is calculated. Next, the host vehicle position information correction unit 11 calculates the position of the host vehicle based on the calculation result of the positional relationship between the host vehicle and the target feature ft and the position information of the target feature ft included in the feature information F. The position information of the subject vehicle is calculated with high accuracy based on the position information (feature information F) of the target feature ft in the traveling direction. The own vehicle position information correcting unit 11 is based on this highly accurate position information of the own vehicle, and includes the current position in the traveling direction of the own vehicle included in the own vehicle position information L acquired by the own vehicle position information acquiring unit 3. Correct the information. As a result, the own vehicle position information acquiring unit 11 acquires the corrected own position information L with high accuracy.

[Lane specific part]
The lane specifying unit 17 acquires lane specifying information J specifying the lane in which the host vehicle is present on the road on which the host vehicle is traveling. The lane specifying unit 17 is proceeding based on the feature information F of the lane marking around the vehicle position on the road where the host vehicle is traveling, and the image recognition result for the lane marking included in the image information G. The lane specifying information J is obtained by performing an operation for specifying the lane. Specifically, for example, the lane specifying unit 17 includes a type (line type such as a solid line, a broken line, and a double line) and an arrangement of the surrounding lines of the host vehicle indicated in the image recognition result by the image recognition unit 6; Based on the form information included in the feature information F of the lane marking around the own position, the lane in which the host vehicle is in progress is specified.

  For example, when the image information G as shown in FIG. 3 is acquired and the map information M around the host vehicle as shown in FIG. 4 is acquired, the lane where the host vehicle exists is a three-lane lane. It can be identified as the middle lane. That is, in the image shown in the image information G shown in FIG. 3, there are broken lane markings on both sides with respect to the center in the width direction of the image, which is the position of the host vehicle, and solid lane markings are further provided on both sides. is there. On the other hand, according to the map information M shown in FIG. 4, the road on which the host vehicle is traveling is three lanes, and the feature information F of the solid line is present on both sides in the width direction of the road. It can be seen that the feature information F of the broken lane markings that divide each lane exists on the center side in the direction. Therefore, the lane specification unit 17 can specify that the lane in which the host vehicle exists is the central lane in the three lanes by comparing these pieces of information.

  Further, the lane specifying unit 17 determines whether or not the lane has been changed based on whether or not the own vehicle straddles the lane line based on the position information of the lane line indicated in the image recognition result, and determines the lane in which the own vehicle is traveling. Identify. The lane identification unit 17 identifies a lane only when it is necessary to identify the lane in which the host vehicle is traveling, that is, when the road on which the host vehicle is traveling has multiple lanes in the traveling direction (one side). Perform the operation. The lane specifying unit 17 then supplies lane specifying information J, which is information for specifying the lane in which the host vehicle is in progress, to the own vehicle position information acquiring unit 3. Thereby, the own position calculation part 3 produces | generates the own position information L containing the lane specific information J regarding the lane in which the own vehicle is advancing as above-mentioned. Therefore, in the present embodiment, the lane identification unit 17 functions as the vehicle position information acquisition unit 18 together with the vehicle position information acquisition unit 3.

[Behavior detection unit]
The behavior detection unit 7 functions as a behavior detection unit that detects the behavior of the host vehicle within a predetermined range from the position where the image recognition unit 6 has succeeded in image recognition of the target feature. As shown in FIG. 1, the behavior detection unit 7 detects the behavior of the host vehicle in response to inputs from various switches and various sensors provided in the host vehicle that function as behavior input means. Examples of the various switches include an air conditioning switch 31, a lighting switch 33, a window switch 35, and an audio operation switch (not shown). The various sensors include, for example, a vibration sensor 37, an illuminance sensor 39, an acceleration sensor (not shown), an accelerator sensor (not shown), a brake sensor (not shown), and the like.

  The air conditioning switch 31 is a switch for switching between setting of air conditioning and heating, introduction of outside air, indoor circulation, and the like. The lighting switch 33 is a switch for switching on / off of a lighting device of a vehicle, setting of a high beam and a low beam, and the like. The window switch 35 is a switch for instructing opening / closing of a window. The vibration sensor 37 is a sensor that detects vibration transmitted to the host vehicle. The detection result is transmitted to, for example, a control device for an active suspension, and the suspension is adjusted to an optimum stiffness. The illuminance sensor 39 is a sensor that detects the brightness outside the host vehicle. The detection result is transmitted to, for example, a control device of the lighting device, and lighting and extinguishing of the lighting device are automatically controlled. The acceleration sensor is a sensor that detects acceleration of the own vehicle and acceleration of deceleration. The accelerator sensor is a sensor that detects an operation amount (accelerator opening) of an accelerator pedal by a driver. The brake sensor is a sensor that detects an operation amount of a brake pedal, a brake depression force, and the like by a driver.

  In the present embodiment, the behavior of the host vehicle detected by the behavior detection unit 7 includes reception of a driver's operation at each part of the host vehicle and the operation of the host vehicle. In addition, the operation of the own vehicle includes the operation of each part of the own vehicle and the entire vehicle caused by the operation of the driver, or the operation of each part of the own vehicle and the entire vehicle caused by an external factor applied to the own vehicle from the outside. included. For example, the operation of various switches such as the air conditioning switch 31, the lighting switch 33, the window switch 35, and the audio operation switch, and the driver's operation detected by various sensors such as an accelerator sensor and a brake sensor are performed in each part of the own vehicle. The behavior detecting unit 7 detects the behavior related to the reception of the driver's operation.

  Further, for example, a change in the traveling direction of the own vehicle detected by the direction sensor 24 by the steering operation of the driver, a change in the acceleration of the own vehicle detected by the acceleration sensor by an operation of the accelerator pedal or the brake pedal by the driver, The behavior of the host vehicle that is detected by various sensors as a result of accepting the driver's operation, such as a change in the gear position of the transmission by a shift operation or an accelerator operation, is a behavior related to the behavior of the host vehicle by the driver's operation. , Detected by the behavior detection unit 7. Further, for example, the operation of the navigation device 1 based on an input to a touch panel or a remote controller provided integrally with the monitor 26 of the navigation device 1 is also a behavior detection unit as a behavior related to the operation of the host vehicle by the driver's operation. 7 is detected. Such operations of the navigation device 1 include, for example, acquisition of traffic jam information, change of map display scale, change of screen display brightness, change of guidance route, and the like based on the operation of the driver.

  Further, for example, the vibration of the own vehicle detected by the vibration sensor 37 by passing through a step or a rough road surface, the change in the acceleration of the own vehicle detected by the acceleration sensor by traveling on a hill, or the direction by traveling on a curve The behavior of the subject vehicle caused by an external factor such as a change in the traveling direction of the subject vehicle detected by the sensor 24 and detected by various sensors or the like is regarded as a behavior related to the behavior of the subject vehicle due to the external factor. Is detected.

  The behavior detection unit 7 detects the behavior of the host vehicle within a predetermined range set based on the host vehicle position information L of the target feature that the image recognition unit 6 has succeeded in image recognition. For example, this behavior is detected when the vehicle turns right or left within a predetermined range after detecting a road marking (diamond mark) indicating the presence of a pedestrian crossing. This behavior is caused by changes in the traveling direction of the host vehicle detected by the direction sensor 24, input from a direction indicator that functions as various switches, input from a steering angle sensor that detects rotation of a steering that functions as various sensors, and the like. Can be detected on the basis. The behavior detection unit 7 associates the detected behavior with the vehicle position information L and outputs it as behavior detection information B.

[Detection result storage means]
Based on the vehicle position information L, the detection result storage unit 8 stores behavior detection information B representing the detection result of the behavior of the vehicle by the behavior detection unit 7 in association with information on the detection position where the behavior is detected. Functions as a detection result storage means. In the present embodiment, the information of the detection position where the behavior is detected is coordinate information representing the detection position of the behavior. The coordinate information representing the detected position of the behavior is derived based on the coordinates of the current position of the host vehicle indicated by the host vehicle position information L when the behavior is detected. Note that the detection result storage unit 8 may store behavior detection information for each individual when means for identifying the individual such as a driver is provided in the host vehicle.

[Learning behavior extraction unit]
The learning behavior extraction unit 9 extracts the behavior of the host vehicle as learning behavior, and outputs the learning behavior attribute information and position information as learning behavior information R associated with the feature information F of the target feature that has been successfully image-recognized. Functions as learning behavior extraction means 29. In the present embodiment, the position information of the learning behavior related to the learning behavior information R is coordinate information representing the position of the learning behavior. The coordinate information representing the position of such learning behavior is information on the detection position of the behavior associated with the behavior detection information B stored in the detection result storage unit 8, here, the detection position of the behavior is as described above. Derived based on the coordinate information to represent. The output learning behavior information R is stored in the learning behavior database 19. This learning behavior database 19 also functions as learning behavior extraction means 29.

  Here, the learning behavior is repeatedly detected based on a plurality of behavior detection information B about the behavior of the same own vehicle, which is stored in the detection result storage unit 8 when the own vehicle passes the same place a plurality of times. It can be the behavior of the host vehicle. For example, there is a behavior in which the direction indicator is operated to the left at a certain point to turn the host vehicle to the left, or a behavior in which the air conditioner is switched from the introduction of outside air to the inside air circulation at a certain point.

  In this case, the learning behavior database 19 can be viewed as each vehicle adaptation database 200 maintained for each vehicle as shown in FIG. As described above, when means for identifying an individual such as a driver is provided, a personal adaptation database may be further configured. On the other hand, the map database 13 and the feature database 15 can be viewed as a common database 100 maintained for roads without depending on each vehicle. As shown in FIG. 5, the common database 100 can be configured to be communicable with the database management center 300 wirelessly or by wire. The map information MC and the feature information FC stored in the map database 13C and the feature database 15C of the database management center 300 are updated as needed. The map database 13 and the feature database 15 of the own vehicle can update the contents of the stored map information M and feature information F to the latest information by communication.

  Further, the learning behavior is the following when the learning behavior extraction unit 9 is connected to be able to communicate with a plurality of vehicles and detects the behavior of the host vehicle due to an external factor applied to the host vehicle from the outside as the behavior of the host vehicle. The behavior can be as follows. FIG. 6 shows an example in which the learning behavior extraction unit 9A of the own vehicle is connected to the learning behavior extraction unit 9B of another vehicle via the database management center 300 so as to be communicable. Of course, the learning behavior extraction units 9 may be able to communicate with each other directly. In such a case, the learning behavior has reproducibility based on a plurality of behavior detection information about the behavior of the vehicle stored in the detection result storage means of each vehicle when the plurality of vehicles pass through the same place. Thus, the behavior of the host vehicle can be detected.

  For example, the behavior is such that a plurality of vehicles detect similar vibrations by the vibration sensor 37 at a certain point. In this case, the vibration may be caused by a road step or the like, and can be regarded as a behavior independent of each vehicle. Therefore, even if each vehicle detects vibrations only once, if a plurality of vehicles detect similar vibrations at almost the same point, the behavior is detected with sufficient reproducibility. be able to.

  In this case, the learning behavior database 19 can be viewed as both the vehicle adaptation database 200 maintained for each vehicle as shown in FIG. 6 and the common database 100 maintained without depending on each vehicle. it can. As shown in FIG. 6, the learning behavior database 19 includes learning behavior information Ra maintained for each vehicle (for each individual), learning behavior information Rb maintained for a road without depending on each vehicle, and the like. Is remembered. The learning behavior information Ra belongs to each vehicle adaptation database 200, and the learning behavior information Rb belongs to the common database 100. As shown in FIG. 6, the common database 100 can be configured to be communicable with the database management center 300 by wireless or wired communication.

  As described above, the map information MC and the feature information FC stored in the map database 13C and the feature database 15C of the database management center 300 are updated as needed. The map database 13 and the feature database 15 of the own vehicle can update the contents of the stored map information M and feature information F to the latest information by communication. Further, the database management center 300 is provided with a learning behavior database 19 </ b> C prepared using a learning behavior extraction unit 9 that is communicably connected to a plurality of vehicles. The learning behavior information RC stored in the learning behavior database 19C is updated as needed. The learning behavior database 19 of the own vehicle can update learning behavior information Rb maintained without depending on each vehicle among the stored data by communication.

[Behavior Prediction Unit]
The behavior prediction unit 10 functions as a behavior prediction unit that predicts the behavior associated with the target feature based on the learning behavior information R. And a prediction result is output with respect to the various control means of the own vehicle, such as the calculating part 12 for navigation. The various control means are a control device in the own vehicle that reproduces the driver's operation, and a control device that optimizes the operation of the own vehicle caused by the driver's operation and external factors. A specific example will be described later.

[Procedure for learning and predicting vehicle behavior]
Hereinafter, the procedure for learning the vehicle behavior and the procedure for predicting the vehicle behavior will be described using the flowcharts of FIGS. 7 and 8. The processing procedure described below is executed by hardware and / or software (program) or both of the above-described functional units. When each of the above functional units is configured by a program, the arithmetic processing device included in the navigation device 1 operates as a computer that executes a vehicle behavior learning program that configures each of the above functional units.

  The own vehicle position information acquisition unit 3 acquires own vehicle position information L indicating the current position of the own vehicle (own vehicle position information acquisition step # 1). Image information G around the host vehicle is acquired by the image information acquisition unit 4 (image information acquisition step # 2). From the feature database 15 in which the feature information F including position information and attribute information about a plurality of target features is stored by the feature information acquisition unit 5, Object information F is acquired (feature information acquisition step # 3). Based on the feature information F, the image recognition unit 6 recognizes the target feature included in the image information G, and the vehicle position information correction unit 11 corrects the vehicle position information L (image recognition). Step # 4). By the above process, when the road on which the host vehicle passes has the feature information F and the target feature is recognized, the host vehicle is in a highly accurate host vehicle position recognition state. When the host vehicle is in such a highly accurate host vehicle position recognition state, the behavior of the host vehicle can be learned as described above. Therefore, it is determined in the application condition determination step # 5 whether or not the vehicle is in a highly accurate vehicle position recognition state. When in the recognition state, the following steps # 11 to # 13 are performed to learn the behavior of the host vehicle.

  Note that if the target feature is traveled for a predetermined distance or longer without being able to recognize the image, the error of the vehicle position information L may increase, and therefore the vehicle is not in a highly accurate vehicle position recognition state. Accordingly, the behavior of the host vehicle within a predetermined range from the small range in which the error of the host vehicle position information L can be regarded as being relatively small, that is, the position where the image recognition unit 6 successfully recognized the target feature is detected. Detected by the unit 7 (behavior detection step # 11). The behavior of the host vehicle detected in the behavior detection step # 11 is at least the operation of the host vehicle caused by the reception of the driver's operation at each part of the host vehicle and the driver's operation or external factors applied to the host vehicle from the outside. Including one. Next, based on the own vehicle position information L, the behavior detection information B representing the detection result of the behavior of the own vehicle by the behavior detection unit 7 is associated with the information on the detection position where the behavior is detected in the detection result storage unit 8. Stored (detection result storage step # 12).

  Then, the learning behavior extraction unit 9 extracts the learning behavior based on the behavior detection information B, and outputs the learning behavior information R (learning behavior extraction step # 13). As described above, the own vehicle is repeatedly detected based on the plurality of behavior detection information B about the behavior of the same own vehicle, which is stored in the detection result storage unit 8 when the own vehicle passes the same place a plurality of times. Are extracted as learning behaviors. Then, the attribute information and position information of the learning behavior are output as learning behavior information R in association with the feature information of the target feature that has been successfully image-recognized. The learning behavior information R is registered in the learning behavior database 19 and is referred to by various control devices in the host vehicle.

  In addition, the learning behavior extracting step # 13 is detected and detected in a plurality of vehicles when the behavior detecting step # 11 detects the operation of the own vehicle due to an external factor applied to the own vehicle from the outside as the behavior of the own vehicle. It is possible to share the behavior detection information B for the behavior associated with the detection position information. In this case, it has reproducibility based on the plurality of behavior detection information B about the behavior of the vehicle, which is stored in the detection result storage unit 8 of each vehicle when the plurality of vehicles pass through the same place. The detected behavior of the host vehicle is extracted as a learning behavior. Then, the attribute information and the position information of the learning behavior are output as learning behavior information R associated with the feature information of the target feature that has succeeded in the image recognition.

  Through the above steps, the behavior of the host vehicle is learned. Hereinafter, a procedure for predicting the behavior of the host vehicle using the learned behavior information R learned will be described. As shown in FIG. 8, the own vehicle position information acquisition step # 1, the image information acquisition step # 2, the feature information acquisition step # 3, and the image recognition step # 4 are performed in the same manner as when learning the behavior of the own vehicle. Is done. According to the above steps, when the road on which the host vehicle passes has the feature information F and the target feature is recognized, the host vehicle is in a highly accurate host vehicle position recognition state. When the host vehicle is in such a highly accurate host vehicle position recognition state, the behavior of the host vehicle can be predicted as described above. Therefore, it is determined in the application condition determination step # 5 whether or not the vehicle is in a highly accurate vehicle position recognition state. When in the recognition state, the following step # 21 is performed, and the behavior of the host vehicle is predicted.

  Based on the learning behavior information R, the behavior prediction unit 10 predicts the behavior associated with the target feature (behavior prediction step # 21). As described above, the behavior of the host vehicle detected in the behavior detection step # 11 is determined by the driver's operation in each part of the host vehicle, the driver's operation, and external factors applied to the host vehicle from the outside. It includes at least one of the operations of the vehicle. Therefore, based on the prediction result in the behavior prediction step # 21, the learning behavior information R is output to the control device in the host vehicle that reproduces the driver's operation. Further, based on the prediction result in the behavior prediction step # 21, the learning behavior information R is output to a control device that optimizes the operation of the host vehicle caused by the driver's operation or external factors.

[Application Example 1]
Hereinafter, specific application examples of the vehicle behavior learning device 2 according to the present invention will be described. FIG. 9 is an explanatory diagram illustrating an example of learning the vehicle behavior based on the trajectory of the vehicle position information L acquired by the vehicle position information acquisition unit 3. In this example, the host vehicle 50 travels on the main road K1 and makes a left turn to the narrow street K3 at the intersection N3 before reaching the large intersection N2 with another main road K2. When the parallel main road K2 and the narrow street K3 are close, that is, when the distance between the intersections N2 and N3 is short, the own vehicle position information acquisition unit 3 determines the current position of the own vehicle 50 indicated by the own vehicle position information L. Map matching may occur on the wrong road. Although the host vehicle 50 is traveling as shown by the solid line in FIG. 9, the current position of the host vehicle 50 indicated by the host vehicle position information L is on the main road K2 as shown by the broken line in FIG. The position of the host vehicle 50 indicated by the host vehicle position information L is displayed on the monitor 26 as the host vehicle position mark 60 in some cases. Such erroneous matching is corrected manually by the driver or automatically by the navigation calculation unit 12, but the vehicle position mark 60 indicating an incorrect position is displayed on the monitor 26 even for a short time. Will be. By using the vehicle behavior learning device 2 of the present invention, such erroneous matching can be suppressed.

  In the example shown in FIG. 9, the driver of the host vehicle 50 operates the direction indicator before reaching the intersection N3, operates the brake to decelerate the host vehicle 50, and performs left turn steering at the intersection N3. Such an operation by the driver to the direction indicator, the brake, and the steering is detected by the behavior detection unit 9 as the behavior of the vehicle. Further, along with such a left turn operation of the host vehicle 50, a change in the traveling direction of the host vehicle 50 detected by the direction sensor 24 is also detected by the behavior detection unit 9 as a behavior of the host vehicle 50 resulting from the driver's operation. Is done. On the other hand, the host vehicle 50 passes through the intersection N1 before approaching the intersection N3. Before and after the intersection N1, there are pedestrian crossings C1 and C2 as features, and these are image-recognized as target features. In this example, it is recognized as a target feature whose position information is an end portion (■ in the figure) on the traveling direction side of the host vehicle 50. Based on one or both of these target features, the bending operation (left turn in this example) of the host vehicle 50 at the intersection N3 is stored in the detection result storage unit 8 as the vehicle behavior. That is, the behavior detection information B representing the detection result of the vehicle behavior is the vehicle position information L corrected by the vehicle position information correction unit 11 based on one or both of the pedestrian crossings C1 and C2 as the target features. Is stored in the detection result storage unit 8 in association with information on the detection position of the behavior based on.

  If the left turn to the narrow street K3 is a route to the home, for example, the host vehicle 50 passes the same route from the main road K1 to the narrow street K3 a plurality of times. Therefore, the detection result storage unit 8 stores a plurality of times of behavior detection information B for the same behavior of the host vehicle 50. The behavior of the host vehicle 50 repeatedly detected in this manner is extracted as a learning behavior by the learning behavior extraction unit 9. The learning behavior attribute information and position information are stored in the learning behavior database 19 as learning behavior information R associated with the feature information of the target feature that has been successfully image-recognized. In this example, the attribute information is information that identifies a left turn at the intersection N3 or an operation of the direction indicator. The position information is coordinate information representing the position of the learning behavior, and is derived based on the information on the detection position of the behavior associated with the behavior detection information B stored in the detection result storage unit 8. In this example, the position indicated in the position information of the learning behavior information R is a position in the intersection N3.

  FIG. 10 is an explanatory diagram illustrating an example in which the prediction result of the vehicle behavior is used in the display of the vehicle position by the navigation calculation unit 12. As described above, a case is considered in which the host vehicle 50 passes the main road K1 in a state where the learning behavior information R of the host vehicle 50 is stored in the learning behavior database 19. When the host vehicle 50 passes through the intersection N1, the image recognition unit 6 recognizes the crosswalks C1 and C2 as target features. The behavior prediction unit 10 acquires the learning behavior information R associated with the pedestrian crossings C1 and C2, and predicts that the host vehicle 50 is highly likely to turn left at the intersection N3. Further, the behavior prediction unit 10 outputs such a behavior prediction result to the navigation calculation unit 12 (an example of navigation calculation means). When the host vehicle 50 turns left at the predicted position, the navigation calculation unit 12 determines that the host vehicle 50 has entered the narrow street K3 instead of the main road K2, and sets the own vehicle position mark 60 to the narrow street. It is displayed on the monitor 26 along with K3. At this time, it is also preferable that the behavior prediction unit 10 outputs a prediction result of such behavior to the vehicle position information acquisition unit 3 to correct the vehicle position information L.

  Thus, by applying the vehicle behavior learning device 2 according to the present invention, the accuracy of the vehicle position display and route guidance in the navigation device 1 can be improved.

[Application Example 2]
In the application example 1 described above, as the behavior of the host vehicle 50, the behavior detection unit 7 detects the operation of the driver accepted in each unit of the host vehicle 50 and the operation of the host vehicle 50 resulting from the operation of the driver. showed that. In the second application example, an example in which an operation of the host vehicle 50 due to an external factor applied to the host vehicle 50 from the outside is detected as a behavior will be described.

  There is a system that controls the damping force of the suspension of the host vehicle 50 based on the road information from the navigation device 1 to improve the steering stability when traveling on a curve or optimally controls the damping force of vibration at a step. These are referred to as a navigation cooperative suspension system because the suspension control and the navigation device 1 are implemented in cooperation. Normally, this navigation cooperative suspension system is implemented mainly using the vehicle position information L acquired by the navigation device 1 through the GPS receiver 23, the direction sensor 24, the distance sensor 25, and the like. However, as described above, since the vehicle position information L includes an error, the suspension control may deviate from the optimum position.

  When the behavior learning apparatus 2 according to the present invention is applied, the behavior of curve running or step passing is detected as behavior detection information B by the suspension vibration sensor 37, the direction sensor 24, the yaw / G sensor, and the like. As described above, this behavior detection information B is associated with the feature information F and stored in the learning behavior database 19 as learning behavior information R. The learning behavior information R can be obtained by extracting the behavior of the own vehicle 50 that is repeatedly detected based on a plurality of behavior detection information B regarding the behavior of the same own vehicle 50.

  On the other hand, this behavior is based on the operation of the host vehicle 50 due to external factors applied to the host vehicle 50 from the outside. For this reason, the dependence on the driver and the vehicle is low, and the dependence on the road itself is high. Therefore, as described with reference to FIG. 6, when a plurality of vehicles pass through the same place, the vehicle has reproducibility based on the plurality of behavior detection information B stored in the detection result storage unit 8 of each vehicle. The behavior of the vehicle detected in this way can also be extracted.

  Based on the learned behavior information R extracted and stored in this manner, behavior prediction unit 10 predicts behavior such as step passing and curve traveling associated with the target feature. Then, the navigation cooperative suspension system is controlled based on the prediction result. Therefore, for example, when there is a step at a specific place on the road, it is possible to predict that a vibration / impact is received when the step is passed, and to perform optimal suspension control. As a result, more precise control is possible than in the prior art. The suspension control device corresponds to a control device that optimizes the operation of the host vehicle 50 in the present invention.

[Application Example 3]
In this application example 3, an example in which the engine of the host vehicle 50, an automatic transmission, and the like are optimally controlled based on road information from the navigation device 1 and reception of a driver's operation will be described. This is referred to as navigation cooperative shift control because the shift control of the automatic transmission or the like and the navigation device 1 are performed in cooperation. For example, when traveling uphill, depending on the driver's preference, the climbing lane may be slowly climbed or the kickdown may be performed at high speed. The behavior detection unit 7 detects, for example, this kick-down operation as behavior detection information B. When this behavior is repeatedly detected based on the same target feature, this driving operation is learned as a driver's habit. That is, this behavior detection information B is stored in the learning behavior database 19 as learning behavior information R in association with the feature information F.

  Based on the stored learning behavior information R, the behavior prediction unit 10 predicts that downshifting is necessary when the image recognition unit 6 recognizes the corresponding target feature. Based on the prediction result, the navigation cooperative shift control system performs optimum shift control in consideration of fuel consumption and the like. This control can be performed for various mechanisms of a so-called power train such as an engine and a transmission. Further, in a hybrid vehicle having an engine and a motor as driving force sources, control can be performed so that the operating state of each driving force source is in an optimum state.

[Application Example 4]
The operation of the sun visor by the driver may be detected as the behavior of the host vehicle 50 related to the reception of the driver's operation. At this time, time information and date information can also be acquired from the GPS receiver 23. Therefore, the learning behavior information R can be extracted and stored together with the time zone, place, and direction in which the driver feels dazzling. When the behavior predicting unit 10 predicts that the driver feels dazzled based on the learned behavior information R, for example, each device performs control such as adjusting the brightness of the monitor 26 or driving the electric sun visor. Implemented by the control unit.

[Application Example 5]
As the behavior of the host vehicle 50 related to the reception of the driver's operation, the operation of the air conditioner by the driver may be detected based on the input from the air conditioning switch 31. For example, a behavior in which a driver who normally uses an air conditioner for introducing outside air repeatedly switches to inside air circulation at a certain point is extracted and stored as learning behavior information R. This is an example of the case where the air conditioning switch 31 is operated so that the exhaust gas of another vehicle does not enter the host vehicle 50 when the driver travels on a highway with a high traffic volume. Based on the learning behavior information R, the behavior prediction unit 10 predicts that the air conditioner is operated. The control unit of the air conditioner automatically switches from outside air introduction to inside air circulation based on the prediction result. If switching to the inside air circulation is delayed and even a slight amount of exhaust gas enters the host vehicle 50, it is uncomfortable, but the automatic switching in this way ensures comfort in the vehicle.

  As described above, according to the present invention, there is provided a vehicle behavior learning device capable of measuring the position of the host vehicle with high accuracy and learning the behavior of a vehicle that is frequently performed at a specific position on a road. can do.

[Other Embodiments]
(1) Although not specifically described in the above embodiment, the feature information F stored in the feature database 15 is irrelevant to the initial feature information stored in advance and the feature information F. It is preferable to use both of the learned feature information learned and stored based on the image recognition result of the feature by the image recognition unit 6. Here, the initial feature information is feature information F about a plurality of features that are prepared and stored in advance in the feature database 15. In order to maintain such initial feature information in all areas where the map information M is maintained, a lot of labor and cost are required. Therefore, the initial feature information is often maintained only for a part of a region such as a metropolitan area or a main road. Therefore, when both the initial feature information and the learned feature information are stored in the feature database 15, the vehicle using the learned feature information in an area where the initial feature information is not maintained. It is preferable to adopt a configuration for learning the behavior of the above.
Here, the learned feature information is learned as follows, for example, and stored in the feature database 15. That is, in order to learn the feature information, the image recognition unit 6 performs image recognition processing of the feature included in the image information G acquired by the image information acquisition unit 4 regardless of the feature information F. If the image recognition of the feature is successful, the recognition position of the feature is obtained, and the recognition position information representing the recognition position is stored in a predetermined feature learning database or the like in a state where the feature can be identified. Remember. Thus, if the host vehicle repeatedly travels on the same road, the same feature is image-recognized a plurality of times, and a plurality of recognized position information for the same feature is stored in the feature learning database. Therefore, the estimated position of the feature is determined by a predetermined estimated position determination unit, and the position information indicating the estimated position of the feature and the image recognition result for the feature are determined by the predetermined learned feature information generation unit. Information associated with the attribute information based on it is generated as learned feature information and stored in the feature database 15.
Of course, the feature information F stored in the feature database 15 may be only the initial feature information or only the learned feature information.

(2) In the above embodiment, the information on the detection position of the behavior associated with the behavior detection information B and the position information on the learning behavior related to the learning behavior information R indicate the detection position of the behavior or the position of the learning behavior. The case where the coordinate information is expressed has been described as an example. However, the embodiment of the present invention is not limited to this, and the information on the detection position of the behavior and the position information on the learning behavior may be information representing the position of the behavior in another manner. Therefore, for example, information indicating the position of the behavior in relation to the target feature that has been successfully image-recognized before detecting the behavior of one or both of the information of the detected position where the behavior is detected and the position information of the learning behavior This is also a preferred embodiment of the present invention. Specifically, for example, the information on the detection position of the behavior is also preferably used as distance information from the position at which the image recognition of the target feature that has been successfully recognized to the detection position of the behavior is successful. If it does in this way, the detection position of a behavior can be appropriately expressed using the distance between the feature and the behavior which shows the relationship with the target feature which succeeded in image recognition. Similarly, the position information of the learning behavior related to the learning behavior information R is also suitable as the distance information from the position of the target feature that has been successfully recognized to the position of the learning behavior. In this way, the position of the learning behavior can be appropriately represented using the distance between the feature and the behavior indicating the relationship with the target feature that has been successfully recognized by the image. In addition, by representing the position of the learning behavior by the distance information with the target feature in this way, when the image recognition of the target feature indicated in the learning behavior information R is successful, the image recognition of the target feature is successful. It is possible to predict the occurrence of learning behavior more accurately based on the determined position.
In addition, one or both of behavior detection position information associated with the behavior detection information B and learning behavior position information related to the learning behavior information R is coordinate information indicating the detection position of the behavior or the position of the learning behavior. And distance information with respect to the target feature.

(3) In the above embodiment, the case where all the configurations of the navigation device 1 including the vehicle behavior learning device 2 are mounted on the host vehicle has been described as an example. However, the embodiment of the present invention is not limited to such a configuration. That is, for example, a part of the configuration excluding the imaging device 21 is installed outside the host vehicle while being connected via a communication network such as the Internet, and transmits and receives information and signals via the network. Thus, configuring the vehicle behavior learning device 2 and the navigation device 1 is also one preferred embodiment of the present invention. Therefore, for example, the map database 13 or the feature database 15 is installed in a server device that is communicably connected to the host vehicle via a wireless communication line or the like, and is acquired from the server device based on the host vehicle position information L. A configuration in which the operation of the navigation device 1 including the vehicle behavior learning device 2 is performed based on the object information F and the map information M is also a preferred embodiment of the present invention.

(4) In the above embodiment, the case where the vehicle behavior learning device 2 includes the behavior prediction unit 10 and outputs the behavior prediction result of the host vehicle 50 to each control unit of the host vehicle 50 will be described as an example. did. However, the configuration of the vehicle behavior learning device 2 according to the present invention is not limited to this, and a configuration not including the behavior prediction unit 10 is also a preferred embodiment of the present invention. For example, the vehicle behavior learning device 2 determines that the behavior of the course change is based on the road shape on the map based on the behavior related to the course change of the host vehicle 50 such as a right turn or a left turn and the road shape included in the map information M. So that the vehicle position information correction means for correcting the vehicle position information L acquired by the vehicle position information acquisition unit 3 is suitable. In this case, the vehicle behavior learning device 2 according to the present invention constitutes a part of the vehicle position recognition device.

  The present invention can be suitably used for a vehicle behavior learning apparatus capable of learning a specific behavior of a vehicle at a specific vehicle position and a vehicle behavior learning program.

The block diagram which shows typically the structural example of the navigation apparatus containing the vehicle behavior learning apparatus which concerns on embodiment of this invention. Explanatory drawing which shows the example of a structure of the information memorize | stored in a map database and a feature database The figure which shows an example of image information The figure which shows an example of the map information around the own vehicle Block diagram schematically showing an example of the structure of a learning behavior database Block diagram schematically showing another example of the structure of the learning behavior database Flow chart showing an example of a procedure for learning vehicle behavior Flow chart showing an example of a procedure for predicting vehicle behavior Explanatory drawing which shows the example which learns a vehicle behavior based on the route guidance of a navigation apparatus Explanatory drawing which shows the example which utilizes the prediction result of a vehicle behavior in the route guidance of a navigation apparatus

Explanation of symbols

1: Navigation device 2: Vehicle behavior learning device 3: Own vehicle position calculation unit (own vehicle position information acquisition means)
4: Image information acquisition unit (image information acquisition means)
5: Feature information acquisition unit (feature information acquisition means)
6: Image recognition unit (image recognition means)
7: Behavior detection unit (behavior detection means)
8: Detection result storage unit (detection result storage means)
9: Learning behavior extraction unit (learning behavior extraction means)
10: Behavior prediction unit (behavior prediction means)
11: Own vehicle position correcting unit (own vehicle position correcting means)
12: Navigation calculation unit (guidance information output means)
13: Map database (map information storage means)
15: Feature database (feature information storage means)
16: Application program (guidance information output means)
17: Lane specific part (own vehicle position information acquisition means)
18: own vehicle position information acquisition means 19: learning behavior database 21: camera, photographing unit 23: GPS receiver 24: direction sensor 25: distance sensor 26: monitor (guidance information output means)
27: Speaker (guidance information output means)
28: Guidance information output unit (guidance information output means)
B: Behavior detection information F, FC: Feature information G: Image information L: Own vehicle position information M, MC: map information R, Ra, Rb, RC: learning behavior information

Claims (11)

Own vehicle position information acquisition means for acquiring own vehicle position information indicating the current position of the own vehicle;
Image information acquisition means for acquiring image information around the host vehicle;
Feature information storage means in which feature information including position information and attribute information about a plurality of target features is stored;
Feature information acquisition means for acquiring the feature information around the host vehicle from the feature information storage means based on the vehicle position information;
Image recognition means for performing recognition processing of the target feature included in the image information based on the feature information;
Behavior detecting means for detecting the behavior of the host vehicle within a predetermined range from a position where the image recognition means has succeeded in image recognition of the target feature;
Detection result storage means for storing behavior detection information representing a detection result of the behavior of the own vehicle by the behavior detection means in association with information on a detection position where the behavior is detected, based on the vehicle position information;
Based on a plurality of behavior detection information on the same behavior stored in the detection result storage means when the host vehicle passes the same place a plurality of times, the behavior of the own vehicle repeatedly detected is extracted as a learning behavior. Learning behavior extraction means for outputting the learning behavior attribute information and position information as learning behavior information associated with the feature information of the target feature that has succeeded in the image recognition ,
The learning behavior extracting means is connected to be able to communicate with a plurality of vehicles, and when the behavior detecting means detects an action of the own vehicle due to an external factor applied to the own vehicle from the outside as a behavior of the own vehicle, the plurality of vehicles are The learning of the behavior of the own vehicle detected with reproducibility based on a plurality of behavior detection information about the behavior of the vehicle stored in the detection result storage means of each vehicle by passing through the same place Vehicle behavior learning device that extracts behavior. The vehicle behavior learning device according to claim 1 , further comprising behavior prediction means for predicting the behavior associated with the target feature based on the learned behavior information. The vehicle behavior learning device according to claim 2 , wherein the behavior prediction unit outputs the prediction result of the behavior to a navigation calculation unit that performs a calculation process for outputting guidance information of the host vehicle. The vehicle behavior learning device according to claim 2 , wherein the behavior prediction unit outputs the prediction result of the behavior to a control device in the host vehicle that reproduces the operation of the driver. The vehicle behavior learning device according to claim 2 , wherein the behavior prediction unit outputs the prediction result of the behavior to a control device that optimizes the operation of the host vehicle. The vehicle behavior learning device according to any one of claims 1 to 5 , wherein the target feature is a road marking provided on a road surface. The said feature information storing means, wherein the feature information, previously stored initial feature information, image recognition of the feature that has been performed independently of the I Ri said feature information in及beauty before Symbol image recognition means The vehicle behavior learning apparatus according to any one of claims 1 to 6 , wherein one or both of learned feature information learned and stored based on a processing result is stored. The information on the detection position where the behavior is detected includes coordinate information representing the detection position of the behavior, and distance information from the position where the image recognition of the target feature that succeeded in the image recognition succeeded to the detection position of the behavior The vehicle behavior learning device according to any one of claims 1 to 7 , including one or both of the following. The learning behavior position information related to the learning behavior information includes one or both of coordinate information representing the position of the learning behavior and distance information from the position of the target feature that has succeeded in the image recognition to the position of the learning behavior. The vehicle behavior learning device according to any one of claims 1 to 8 , further comprising: The vehicle behavior learning device according to any one of claims 1 to 9 ,
Map information storage means for storing map information;
A plurality of application programs that operate with reference to one or both of the learning behavior information and the map information output by the vehicle behavior learning device;
Guidance information output means that operates according to the application program and outputs guidance information;
A navigation device comprising: Own vehicle position information acquisition step of acquiring own vehicle position information indicating the current position of the own vehicle;
An image information acquisition step of acquiring image information around the host vehicle;
A feature that acquires the feature information around the host vehicle based on the host vehicle position information from a feature information storage unit that stores feature information including position information and attribute information about a plurality of target features. Information acquisition process;
Based on the feature information, an image recognition step for performing recognition processing of the target feature included in the image information;
A behavior detection step of detecting the behavior of the host vehicle within a predetermined range from a position at which image recognition of the target feature by the image recognition means is successful;
Based on the vehicle position information, a detection result for storing behavior detection information representing a detection result of the behavior of the vehicle by the behavior detection unit in a detection result storage unit in association with information on a detection position where the behavior is detected Memory process;
Based on a plurality of behavior detection information about the behavior of the same own vehicle stored in the detection result storage means when the own vehicle passes the same place a plurality of times, the behavior of the own vehicle repeatedly detected as a learning behavior Extracting and outputting the learning behavior attribute information and the position information of the learning behavior as learning behavior detection information associated with the feature information of the target feature that has succeeded in the image recognition, and causing the computer to execute ,
In the learning behavior extracting step, when detecting the operation of the host vehicle due to an external factor applied to the host vehicle from the outside as the behavior of the host vehicle by the behavior detecting step, the plurality of vehicles communicate with each other, Based on a plurality of behavior detection information about the vehicle behavior stored in the detection result storage means of each vehicle by passing, the behavior of the host vehicle detected with reproducibility is extracted as the learning behavior vehicle behavior learning program to be.

Images