JP2016038757A - Traffic light recognition apparatus and traffic light recognition method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the accuracy of detecting traffic lights when two or more traffic lights are to be included in an image.SOLUTION: A traffic light recognition apparatus (100) includes an imaging section 11 which images the surroundings of a vehicle to obtain an image, detects a position of the vehicle (12), and estimates a relative position of a traffic light which the vehicle is to follow, with respect to the vehicle, from the position of the vehicle and position information of the traffic light located around the vehicle (13). A detection area is set in the position of the traffic light on the image estimated from the relative position of the traffic light (14). The traffic light is detected from the detection area set on the image (15). For each of the detection areas, probability of falsely detecting the traffic light is evaluated (16). When two or more detection areas are set, the traffic light is detected from a detection area having the lowest probability of falsely detecting the traffic light.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a traffic signal recognition apparatus and a traffic signal recognition method.

  2. Description of the Related Art Conventionally, a vehicle outside recognition device that recognizes a traffic light or a stop sign from an image obtained by capturing the traveling direction of a vehicle is known (Patent Document 1). When there are a plurality of traffic lights or stop signs in the image, in Patent Document 1, the arrangement order along the traveling direction of the vehicle is determined, and according to the traffic lights or stop signs present at the closest position to the vehicle Control the running state of the vehicle.

JP 2007-257299 A

  However, in Patent Document 1, when selecting a traffic signal serving as a reference for running control from among a plurality of traffic signals, the distance from the vehicle to the traffic signal is taken into consideration, but in consideration of the possibility of erroneous detection of each traffic signal. Not. An object that is similar in color, brightness, or shape to a traffic light may be erroneously detected as a traffic light.

  The present invention has been made in view of the above problems, and provides a traffic signal recognition apparatus and a traffic signal recognition method capable of increasing the detection accuracy of traffic signals when two or more traffic signals are predicted to appear in an image. It is an object.

  A traffic light recognition apparatus according to an aspect of the present invention includes an imaging unit that captures an image of a vehicle to acquire an image, detects a self-position of the vehicle, and uses position information and a self-position of a traffic light around the vehicle. The relative position of the traffic light to be followed by the vehicle is estimated. A detection area is set at the position of the traffic light on the image expected from the relative position of the traffic light, and the traffic light is detected from the detection area set on the image. Then, the ease of erroneous detection of the traffic signal is evaluated for each of the detection areas. When two or more detection areas are set, the traffic signal is detected from the detection area where the erroneous detection of the traffic signal is the lowest.

  According to the traffic signal recognition device and the traffic signal recognition method, by detecting the traffic signal from the detection areas that are least susceptible to erroneous detection of the traffic signal among two or more detection areas, it is possible to suppress erroneous detection of the traffic signal. it can. Therefore, the detection accuracy of the traffic light can be increased.

FIG. 1 is a block diagram illustrating information input to and output from the traffic signal recognition apparatus 100 according to the embodiment. FIG. 2 is a block diagram showing the configuration and data flow of the traffic signal recognition apparatus 100 according to the embodiment. FIG. 3 is a flowchart showing an example of a signal recognition method using the traffic signal recognition apparatus 100 shown in FIG. FIG. 4 is a flowchart showing a detailed procedure of step S13 of FIG. FIG. 5 is a diagram illustrating an example of a travel route for explaining an example of a false detection degree evaluation method by the false detection degree evaluation unit 16. FIG. 6 is a diagram illustrating an example of an image IMG captured by the imaging unit 11 mounted on the vehicle VC in the travel route example of FIG. FIG. 7 is a diagram showing detection regions RG1 and RG2 that are set in the image IMG of FIG. 6 and in which the traffic lights TS1 and TS2 are predicted to be captured. FIG. 8A is a diagram showing a detection region RG1 cut out from the image IMG in FIG. 7, and FIG. 8B is a diagram showing a detection region RG2 cut out from the image IMG in FIG. FIG. 9A and FIG. 9B are tables showing examples of the degree of false detection calculated from the color information of each detection region (RG1, RG2). FIG. 10A and FIG. 10B are tables showing examples of the degree of false detection calculated from the signal shape of the detection region (RG1, RG2). FIGS. 11A and 11B are tables showing examples of the degree of false detection calculated from the luminance distribution of the detection regions (RG1, RG2). FIG. 12A and FIG. 12B are tables showing examples of false detection degrees calculated from combinations of color information, signal shape, and luminance distribution of the detection regions (RG1, RG2).

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. The same members are denoted by the same reference numerals and the description thereof is omitted.

  With reference to FIG. 1, the information input / output to / from the traffic signal recognition apparatus 100 according to the embodiment will be described. The traffic signal recognition apparatus 100 detects a traffic signal installed around a road on which the vehicle travels from an image captured by an imaging unit (camera) mounted on the vehicle.

  Map information D02, landmark information D01, and camera information D03 are input to the traffic signal recognition apparatus 100. The map information D02 includes traffic signal position information associated in advance between the real environment and the map. The landmark information D01 is used to calculate the vehicle's own position in the real environment. The landmark includes a feature (ground landmark) provided on the ground and a GPS satellite that transmits a GPS signal that can be received by the vehicle. In the embodiment, a ground landmark will be described as an example. The landmark information D01 includes, for example, ground landmark position information. The camera information D03 is used to extract a video around the vehicle (for example, forward) from the imaging unit. The traffic signal recognition apparatus 100 outputs a traffic signal detection result as traffic signal information D04 based on the information D01 to D03.

  With reference to FIG. 2, the configuration and data flow of the traffic signal recognition apparatus 100 according to the embodiment will be described. The traffic signal recognition apparatus 100 includes an imaging unit 11, a self-position detection unit 12, a traffic signal position estimation unit 13, a detection area setting unit 14, a traffic signal detection unit 15, and an erroneous detection degree evaluation unit 16.

  The imaging unit 11 is mounted on a vehicle and captures an image of the surroundings of the vehicle. The imaging unit 11 is a camera including a solid-state imaging device, for example, a CCD and a CMOS, and acquires an image that can be processed. The imaging unit 11 sets the angle of view of the lens and the vertical and horizontal angles of the camera based on the camera information D03, and outputs the acquired image as image data D08.

  The self position detector 12 detects the self position of the vehicle based on the landmark information D01. The landmark information D01 is, for example, information on the relative position of the ground landmark (store, famous place, sightseeing spot) with respect to the vehicle detected by a vehicle-mounted camera or a sensing means such as a laser radar. In the map information D02, position information of ground landmarks is registered in advance. The self-position of the vehicle can be detected by comparing the landmark information D01 with the information on the relative position of the ground landmark. Here, the “position” includes coordinates and orientation. Specifically, the position of the ground landmark includes the coordinates and orientation of the ground landmark, and the position of the vehicle includes the coordinates and orientation of the vehicle. The self-position detector 12 outputs the coordinates (x, y, z) in the reference coordinate system and the posture (pitch, yaw, roll) that is the rotation direction of each coordinate axis as the self-position information D05.

  For example, the self-position detecting unit 12 detects the initial position of the vehicle using the landmark information D01. The initial position is a position of the vehicle that can be directly obtained from the landmark information D01, that is, a coordinate and a posture. Then, the vehicle's own position is calculated by cumulatively adding the movement amount of the vehicle to the initial position. The self-position detection unit 12 can estimate the amount of movement of the vehicle per unit time, that is, the amount of change in coordinates and posture, using odometry, a radar device, a gyro sensor, a yaw rate sensor, and a rudder angle sensor.

  The traffic signal position estimation unit 13 estimates the relative position of the traffic signal with respect to the vehicle from the map information D02 and the self-position information D05. In the map information D02, traffic signal position information (coordinate information) is registered in advance. The relative coordinates of the traffic signal with respect to the vehicle can be obtained from the coordinates of the traffic signal and the coordinates and posture of the vehicle. The traffic signal position estimation unit 13 outputs the estimated relative coordinates of the traffic signal as relative position information D06. The traffic signal whose relative position is estimated by the traffic signal position estimation unit 13 is a traffic signal that presents a signal to the vehicle, in other words, a traffic signal that the vehicle should follow. For example, the traffic signal position estimating unit 13 extracts the intersection where the traffic signal is installed that is the nearest intersection where the vehicle will enter from now on as the “nearest intersection”. At least one traffic signal that is present at the “closest intersection” and that the vehicle of the self-location information should follow is identified as the “closest traffic signal”. Then, the relative position of the “nearest signal” with respect to the vehicle is output as relative position information D06.

  The detection area setting unit 14 sets the detection area of the traffic light in the image from the relative position of the traffic light. Since the imaging unit 11 is fixed to the vehicle, the position on the image where the traffic light is expected to be captured can be specified in the image if the angle of view and direction taken by the imaging unit 11 are determined. . The detection area setting unit 14 sets a detection area based on the position on the image where the traffic light is predicted to be captured. The detection area setting unit 14 outputs a detection area in which the traffic light is predicted to appear as detection area information D09.

  The error detection degree evaluation unit 16 evaluates the likelihood of erroneous detection of a traffic light, that is, the possibility of erroneous detection of a signal other than a traffic light as a traffic light (error recognition degree) for each detection region. For example, the erroneous detection degree evaluation unit 16 extracts at least one of color, shape, and luminance from the image data D08 and the detection area information D09, and evaluates the erroneous detection degree using the extracted information. The evaluation result for each detection area is output as erroneous detection information D10.

  Specifically, the false detection degree evaluation unit 16 can evaluate the ease of false detection of a traffic light according to the ratio of the color area similar to the color of the signal lamp included in the traffic light in the detection area. The more the color areas similar to the color of the signal lamp are distributed in the detection area, the higher the false detection degree is evaluated. The false detection degree evaluation unit 16 can evaluate the ease of false detection of a traffic light according to the number of shapes similar to the traffic light included in the detection region. The more the shape of the signal lamp (circular) or the shape similar to the shape of the frame of the traffic light is included in the detection area, the higher the false detection degree is evaluated. And the misdetection degree evaluation part 16 can evaluate the ease of misdetection of a signal apparatus according to the ratio for which the area | region of the brightness | luminance similar to the signal lamp which a signal apparatus has occupies in a detection area. The higher the ratio of the luminance area similar to the signal lamp included in the traffic light to the detection area, the higher the false detection degree is evaluated.

  The traffic light detection unit 15 detects a traffic signal from the detection area set by the detection area setting unit 14. When two or more detection areas are set in one image, the traffic light detection unit 15 detects a traffic light from the detection areas with the lowest possibility of erroneous detection of the traffic light (degree of erroneous detection). Thereby, the false detection of a traffic signal can be suppressed.

  The traffic signal detector 15 can continuously detect the traffic signal when it can detect the traffic signal from the detection region where the erroneous detection of the traffic signal is the lowest. Thereby, it is possible to stably detect the traffic light.

  When the traffic signal detection unit 15 cannot detect a traffic signal from the detection area where the signal error is least likely to be detected, the traffic signal detection unit 15 detects the traffic signal from the detection area where the traffic signal detection error is the next lowest. Also good. Thereby, a traffic signal can be detected at an early stage.

  The traffic light detection unit 15 can cut out the detection area set by the detection area setting unit 14 from the image and detect the traffic signal from the cut out detection area. Alternatively, the detection area setting unit 14 may control the attitude of the imaging unit 11 with respect to the vehicle and the angle of view of the imaging unit 11 so that an enlarged image obtained by enlarging the set detection area can be acquired. In this case, the traffic light can be detected from the enlarged image acquired by the controlled imaging unit 11, that is, the detection area. Thereby, since the resolution of a detection area increases, the recognition accuracy of a traffic light improves. Using the imaging unit 11 having a pan / tilt zoom mechanism, an enlarged image may be captured by narrowing down the entire image space to the detection area.

  Next, an example of a signal recognition method using the traffic signal recognition device 100 shown in FIG. 2 will be described with reference to FIG. The flow shown in FIG. 3 is repeatedly performed based on a predetermined cycle.

  First, in step S01, the imaging unit 11 captures an image around the vehicle based on the camera information D03 and acquires an image. In step S03, the self-position detecting unit 12 detects the self-position of the vehicle using the landmark information D01, and outputs the detected self-position as self-position information D05.

  Proceeding to step S05, the traffic signal position estimation unit 13 estimates the relative position of the traffic signal with respect to the vehicle from the map information D02 and the self-position information D05, and outputs the estimated relative position of the traffic signal as relative position information D06.

  Proceeding to step S07, the detection area setting unit 14 estimates a position on the image where the traffic light is predicted to be captured in the image from the relative position of the traffic light, and sets the detection area. The set detection area is output as detection area information D09.

  In step S09, the detection area setting unit 14 determines whether two or more detection areas are set. If two or more detection areas are set (YES in S09), the process proceeds to step S11. If one detection area is set (NO in S09), the process proceeds to step S13.

  In step S <b> 11, the erroneous detection degree evaluation unit 16 evaluates the erroneous detection degree using at least one information of color, shape, and luminance in the detection region. Thereafter, the process proceeds to step S13. The evaluation result for each detection area is output as erroneous detection information D10.

  In step S <b> 13, the traffic signal detection unit 15 detects a traffic signal from the detection area set by the detection area setting unit 14. When two or more detection areas are set in one image, the traffic signal detection unit 15 detects a traffic signal from the detection areas with the lowest false detection level of the traffic signal.

  A detailed procedure of step S13 in FIG. 3 will be described with reference to FIG. Here, a case where N detection areas (N is a natural number of 2 or more) are set in step S07 in FIG. 3 will be described. First, in step S21, image data of each of the N detection areas is acquired. The image data of the detection area may be acquired by cutting out an image, or an enlarged image may be acquired by controlling the attitude and angle of view of the imaging unit 11. Proceeding to step S23, false detection information D10 for each detection region is acquired. Proceeding to step S25, an attempt is made to detect a traffic light from a detection area having the lowest error detection degree among the N detection areas.

  If the traffic signal can be detected (YES in S25), the process proceeds to step S27, and the traffic signal information D04 is output. Thereafter, without changing the rank (i) of the false detection degree (step S29), the process returns to step S25. Thereby, when the detection of the traffic signal is successful (YES in S25), the traffic signal can be continuously detected from the same detection area. “Order (i)” is a natural number of 1 or more and N or less, and the lower the false detection degree, the smaller the value. That is, the rank (i) of the detection area with the lowest false detection degree is 1, and the rank (i) of the detection area with the highest false detection degree is N. The initial setting is i = 1.

  On the other hand, if the traffic signal cannot be detected (NO in S25), the process proceeds to step S31, and it is determined whether or not the current rank (i) is N, that is, whether the false detection degree is the highest value. If the current rank (i) is not N (NO in S31), a detection area that has not yet been attempted to be detected remains. Therefore, the current rank (i) is incremented (increased) by 1 (step S35), that is, the current rank is lowered by 1, and the process returns to step S25. On the other hand, if the current rank (i) is N (YES in S31), detection is attempted for all detection areas. Therefore, the current rank (i) is set to the highest rank again (step S33), and the process returns to step S25.

  As described above, the traffic signal detection unit 15 tries to detect traffic signals in order from the detection region where the erroneous detection of the traffic signal is low, and if it can be detected, the traffic signal detection unit 15 continuously detects the traffic signal from the detection region. When the traffic signal cannot be detected, it tries to detect the traffic signal from the detection area where the signal error is less likely to be detected.

  Next, an example of the evaluation method of the erroneous detection degree by the erroneous detection degree evaluation unit 16 will be described. As shown in FIG. 5, the vehicle VC on which the imaging unit 11 is mounted travels in the own lane WL of the road RD that faces the vehicle, and the oncoming vehicle OP travels in the oncoming lane PL. Both the vehicle VC and the oncoming vehicle OP are traveling toward the intersection Crs that intersects the road CD.

  Traffic lights TS1 and TS2 are provided at the intersection Crs. The traffic light TS1 is located on the own lane WL side, and the traffic light TS2 is located on the opposite lane PL side. Further, a circular sign MA1 is provided on the road shoulder on the own lane WL side, and circular signs MA2 and MA3 and square signs BD1 and BD2 are provided on the road shoulder on the opposite lane PL side. The colors of the signs MA1, MA2, MA3 are red, which is similar to a red signal light. The sign BD1 is lit in red similar to a red signal light. The sign BD2 is lit in blue similar to a blue signal light. The color of the vehicle body of the oncoming vehicle OP is red, similar to a red signal light. The intersection Crs corresponds to the “nearest intersection”, and the traffic lights TS1 and TS2 correspond to the “nearest traffic signal”.

  FIG. 6 shows an example of an image IMG captured by the imaging unit 11 mounted on the vehicle VC in the example of the traveling path of FIG. Of the signal lights included in the traffic lights TS1 and TS2, the red signal lights RP1 and RP2 are turned on, and the blue and yellow signal lights are turned off. As shown in FIG. 7, the detection area setting unit 14 sets detection areas RG1 and RG2 in which the traffic lights TS1 and TS2 are predicted to appear in the image IMG. The detection region RG1 is a region including the own lane and the road shoulder on the own lane side, and the detection region RG2 is a region including the opposite lane and the road shoulder on the opposite lane side.

  As shown in FIG. 8A, the traffic signal TS1 and the sign MA1 are shown in the detection region RG1. As shown in FIG. 8B, the traffic signal TS2, signs MA2, MA3, signboards BD1, BD2 are shown in the detection region RG2.

(Error detection evaluation method based on color information)
The false detection degree evaluation unit 16 evaluates the degree of possibility that a signal other than the traffic light is erroneously detected as a traffic light according to the color distribution similar to the color of the traffic light included in the traffic light. For example, the false detection degree evaluation unit 16 may detect that the “color distribution included in the blue signal”, the “color distribution included in the red signal”, and the “color distribution included in the yellow signal” are the detection regions RG1, Calculate how much is contained in RG2. An example of the calculation result is shown in FIGS. 9A and 9B.

  In the detection region RG1, a red sign MA1 is present in addition to the signal lamp RP1 that is lit red. For this reason, the false detection degree evaluation unit 16 calculates “the color distribution included in the red signal” as 2, calculates “the color distribution included in the blue signal” as 0, and “includes in the yellow signal”. The color distribution is calculated as 0. The false detection degree of the detection region RG1 is calculated as 2 by adding these numerical values.

  In the detection region RG2, there are red signs MA2 and MA3, a signboard BD1 that lights red, and an oncoming vehicle OP that has a red vehicle body, in addition to the signal lamp RP2 that lights red. Further, there is a signboard BD2 that lights up in blue. For this reason, the false detection degree evaluation unit 16 calculates “the color distribution included in the red signal” as 11, calculates “the color distribution included in the blue signal” as 2, and “included in the yellow signal”. The color distribution is calculated as 0. The false detection degree of the detection region RG2 is calculated as 13 by adding these numerical values. Therefore, it can be determined that the degree of erroneous detection based on color information is higher in the detection region RG2 than in the detection region RG1.

(Error detection evaluation method based on signal shape)
The false detection degree evaluation unit 16 evaluates how much there is a possibility that a signal other than the traffic light is erroneously detected as a traffic light depending on the number of shapes similar to the traffic light included in the detection region. The “shape similar to a traffic light” may be any of a round signal light shape, a combination of a square traffic light frame shape and a round signal light shape, or a shape specific to a traffic light. As an example of “a shape peculiar to a traffic light”, there is a combination of three round shapes arranged in a line in the vertical direction or the horizontal direction.

  In the detection region RG1, there is a round sign MA1 other than the three signal lights included in the traffic light TS1. Therefore, the false detection degree evaluation unit 16 calculates the false detection degree based on the “signal shape” as 4. On the other hand, in the detection region RG2, round signs MA2 and MA3 exist in addition to the three signal lights included in the traffic light TS2. For this reason, the false detection degree evaluation unit 16 calculates the false detection degree based on the “signal shape” as 10. Therefore, it can be determined that the detection error level due to the signal shape is higher in the detection region RG2 than in the detection region RG1.

  Note that the number of three combinations selected from all the round shapes included in the detection region can also be calculated as the false detection degree based on the “signal shape”. Furthermore, when four or more round shapes are arranged in the vertical direction or the horizontal direction in the detection region, the number of combinations of three round shapes selected from the four or more round shapes is set as follows: It can also be calculated as a false detection degree based on the “signal shape”.

(Error detection evaluation method using luminance distribution)
The false detection degree evaluation unit 16 evaluates the degree of possibility that a signal other than the traffic light is erroneously detected as a traffic light according to the distribution of luminance similar to the traffic light included in the traffic light in the detection region. Note that the hue need not be taken into account in this evaluation method. In the detection region RG1, there is no similar luminance region other than the signal lamp RP1 that lights red. For this reason, the erroneous detection degree evaluation unit 16 calculates 1 as the erroneous detection unit based on the “luminance distribution”. On the other hand, in the detection region RG2, there are signboards BD1 and BD2 that are lit with similar luminance other than the signal lamp RP1 that is lit red, and the headlamps of the oncoming vehicle OP. For this reason, the erroneous detection degree evaluation unit 16 calculates 3 as the erroneous detection unit based on the “luminance distribution”. Therefore, it can be determined that the detection error level due to the luminance distribution is higher in the detection region RG2 than in the detection region RG1.

(Error detection evaluation method based on combination of color, shape, and brightness)
You may evaluate how much there is a possibility of misdetecting things other than a traffic light as a traffic light by combining at least any two information of a color, a shape, and a brightness | luminance. FIG. 12 shows an example of a false detection evaluation method combining three pieces of information of color, shape, and luminance. Here, as shown in the equation (1), a numerical value obtained by adding the above-described “false detection degree by color information”, “false detection degree by signal shape”, and “false detection degree by luminance distribution” for each detection region, Calculated as the degree of false detection. In the equation (1), F _CN , F _FN , and F _IN indicate “a false detection degree by color information”, “a false detection degree by signal shape”, and “a false detection degree by luminance distribution”, and _GN is calculated. The detected false positive degree is shown.

G _N = F _CN + F _FN + F _IN (1)

Accordingly, as shown in FIG. 12 (a), false detection of _{G A} detection region RG1 is calculated to be 4. As shown in FIG. 12 (b), false detection of _{G B} of the detection region RG1 is calculated to be 26. Therefore, also in this case, it can be determined that the detection error level due to the luminance distribution is higher in the detection region RG2 than in the detection region RG1.

Alternatively, (2) as shown in the expression "degree of detection error due color information 'described above, each of the" degree of detection error due signal shape "and" detection of erroneous luminance distribution ", weighted (W _C, W _F , W _I ) may be added.

_{G N = W C × F CN} + W F × F FN + W I × F IN (2)

By changing the weights (W _C , W _F , W _I ) in accordance with the traveling state of the vehicle, appropriate signal recognition processing can be performed in accordance with the traveling state. For example, when driving at night or in a tunnel, the ambient light is weak, so it may be difficult to capture not only a signal light during lighting but also a lighted signal light and a frame of a traffic light adjacent to this. . In such a case, it is difficult to recognize the arrangement of a plurality of round shapes, the shape of the frame of the traffic light, and the like. Therefore, than the weight of the "degree of detection error due signal shape" (W _F), increasing the weight _{(W C,} W _I) of the "degree of detection error due color information" and the "degree of detection error due luminance distribution". Accordingly, it is possible to perform appropriate signal recognition processing while traveling at night or traveling through a tunnel. On the other hand, in daytime and driving situations with strong ambient light, the color distribution and luminance distribution similar to signal lights may increase. In such a case, the color and luminance weights may be reduced (W _C , W _I ) and the shape weights (W _F ) may be increased.

  As described above, according to the present invention, the following effects can be obtained.

  When the detection region setting unit 14 sets two or more detection regions (RG1, RG2) in one image, the false detection degree evaluation unit 16 makes it easy to falsely detect a traffic light for each of the detection regions (false detection degree). ) And the traffic signal detection unit detects the traffic signal from the detection area where the false detection of the traffic signal is the lowest. Thereby, the false detection of a traffic signal can be suppressed. Therefore, it is possible to detect the traffic signal from the image with high accuracy. In addition, since an attempt is made to detect a traffic light from the detection area of the image, the calculation load is reduced.

  The false detection degree evaluation unit 16 can evaluate the ease of false detection of the traffic light according to the ratio of the color area similar to the color of the signal lamp included in the traffic light in the detection area. It is possible to suppress erroneous recognition of a color area similar to a signal light as a traffic light.

  The false detection degree evaluation unit 16 can evaluate the ease of false detection of a traffic light according to the number of shapes similar to the traffic light included in the detection region. It is possible to suppress erroneous recognition of a region having a shape similar to that of a traffic light as a traffic light.

  The false detection degree evaluation unit 16 can evaluate the ease of false detection of a traffic light in accordance with the ratio of the brightness area similar to the signal lamp included in the traffic light in the detection area. It is possible to suppress erroneous recognition of a region having a luminance similar to that of a signal lamp as a traffic light.

  As shown in FIG. 4, when the traffic signal detector 15 can detect the traffic signal from the detection area where the erroneous detection of the traffic signal is the lowest (YES in S25), the traffic signal detector 15 continuously detects the traffic signal. Thereby, it is possible to stably detect the traffic light.

  In some cases, it is possible to detect a traffic light first from a detection area having the next lowest likelihood of erroneous detection, rather than waiting for a traffic light to be detected from a detection area having the lowest traffic light misdetectability. Therefore, as shown in FIG. 4, when the traffic light detection unit 15 cannot detect the traffic signal from the detection area where the erroneous detection of the traffic light is the lowest (NO in S25), The traffic light is detected from the low detection area. Thereby, a traffic signal can be detected at an early stage. Also, the calculation load is reduced.

  The detection area setting unit 14 may control the posture of the imaging unit 11 relative to the vehicle and the angle of view of the imaging unit 11 so that an enlarged image obtained by enlarging the detection areas (RG1, RG2) can be acquired. Since the resolution of the detection area is higher than when the detection area is cut out from the image, the recognition accuracy of the traffic light is improved.

  Although the contents of the present invention have been described with reference to the embodiments, the present invention is not limited to these descriptions, and it is obvious to those skilled in the art that various modifications and improvements can be made.

DESCRIPTION OF SYMBOLS 11 Image pick-up part 12 Self-position detection part 13 Traffic signal position estimation part 14 Detection area setting part 15 Traffic signal detection part 16 False detection degree evaluation part 100 Traffic signal recognition apparatus D02 Map information RG1, RG2 Detection area

Claims (8)

An imaging unit mounted on a vehicle and capturing an image of the surroundings of the vehicle;
A self-position detector for detecting the self-position of the vehicle;
A traffic signal position estimation unit that estimates a relative position of the traffic signal to be followed by the vehicle from map information including positional information of traffic signals around the vehicle and the self-position;
A detection area setting unit for setting a detection area at the position of the traffic light on the image expected from the relative position of the traffic light;
A traffic light detector for detecting the traffic light from a detection area set on the image;
An error detection degree evaluation unit that evaluates the ease of erroneous detection of the traffic light for each of the detection areas, and
The traffic signal detection unit detects the traffic signal from the detection area where the detection error is least likely when the detection area is set to 2 or more,
A traffic light recognition apparatus characterized by the above.   The false detection degree evaluation unit evaluates the ease of false detection of the traffic light according to a ratio of a color region similar to the color of a signal light included in the traffic light in the detection region. Item 2. A traffic signal recognizing device according to item 1.   The said misdetection degree evaluation part evaluates the ease of misdetection of the said signal apparatus according to the number of the shapes similar to the said signal apparatus contained in the said detection area, The Claim 1 or 2 characterized by the above-mentioned. Traffic light recognition device.   The error detection degree evaluation unit evaluates the ease of erroneous detection of the traffic light according to a ratio of a brightness area similar to a signal lamp included in the traffic light in the detection area. The traffic signal recognition apparatus as described in any one of -3.   The said traffic signal detection part detects the said traffic signal continuously, when the said traffic signal can be detected from within the said detection area where the misdetection of the said traffic signal is the lowest, The said traffic signal is detected. The traffic signal recognition apparatus as described in any one of Claims.   When the traffic light detection unit fails to detect the traffic signal from the detection area where the traffic signal is least likely to be erroneously detected, the traffic light detection unit detects from the detection area where the traffic signal is less likely to be erroneously detected. The traffic light recognition apparatus according to claim 1, wherein the traffic light is detected.   The said detection area setting part controls the attitude | position with respect to the said vehicle of the said imaging part, and the angle of view of the said imaging part so that the enlarged image which expanded the said detection area can be acquired. The traffic signal recognition apparatus as described in any one of Claims. Using an imaging unit mounted on the vehicle, image the surroundings of the vehicle to obtain an image,
Detecting the position of the vehicle,
From the map information including the position information of traffic lights around the vehicle and the self-position, the relative position of the traffic lights to the vehicle to be followed by the vehicle is estimated,
Set the detection area to the position of the traffic light on the image expected from the relative position of the traffic light,
Detecting the traffic light from a detection area set on the image;
A traffic light recognition method for evaluating ease of erroneous detection of the traffic light for each of the detection areas,
When the detection area is set to 2 or more, the traffic signal is detected from the detection area with the lowest possibility of erroneous detection of the traffic light.
A traffic light recognition method characterized by the above.

Images