JP4096932B2 - Vehicle collision time estimation apparatus and vehicle collision time estimation method - Google Patents

Description

  The present invention relates to a vehicle collision time estimation device and a vehicle collision time estimation method that are mounted on a vehicle and estimate a collision time with an object.

  Conventionally, a horizontal or vertical edge of an object with a possibility of collision is detected from two temporally different vehicle surrounding images, and the optical flow of the detected edge is calculated by a correlation method. A vehicle collision time estimation device that estimates a collision time with an object based on a flow is known (see, for example, Patent Document 1). And according to such a collision time estimation device for vehicles, it can make a driver take collision avoidance action to a subject by issuing a warning based on the estimated collision time.

Note that the correlation method is a block matching technique that divides an image into a plurality of areas and finds a similar area from temporally continuous images. In this technique, the attention area I in the frame at time t is used. For (x, y, t) and a region I (x + u, y + v, t + 1) corresponding to a region of interest I (x, y, t) in a frame temporally continuous with the frame at time t, ΣΣ {I ( x, y, t) −I (x + u, y + v, t + 1)} ² and ΣΣ | I (x, y, t) −I (x + u, y + v, t + 1) | It is detected as an optical flow of the object between successive frames.
JP 11-353565 A

  However, when optical flow is detected using the correlation method, a correlation value is calculated around the region of interest, and the calculation is performed to detect the correlation value of the in-region luminance between two temporally different frames. The amount increases and the collision time cannot be calculated easily. In addition, since the edge speed in the image is calculated after detecting the corresponding position in the image, the edge positioning accuracy in the image directly affects the collision time estimation accuracy. That is, if an error is included in edge position detection, the error is included in the edge movement amount.

  The present invention has been made to solve the above-described problems, and an object of the present invention is to make it possible to easily calculate a collision time and obtain a robust output with respect to a position detection error. The object is to provide a time estimation device and a vehicle collision time estimation method.

  In order to solve the above-described problem, the vehicle collision time estimation device and the vehicle collision time estimation method according to the present invention normalize the edge width of the edge image, and the edge image is detected for the normalized edge image. The count value corresponding to the selected position is incremented, the count value corresponding to the position where the edge image is not detected is initialized, and the moving direction and moving speed of the edge image are calculated based on the inclination of the count value. The collision time with the object is calculated using the position and the moving speed of the edge image.

  According to the vehicle collision time estimation device and the vehicle collision time estimation method according to the present invention, the moving speed and the collision time of the edge image can be calculated without performing a block matching process such as template matching. It is possible to easily calculate the time and to obtain an output robust to the position detection error.

  Hereinafter, the configuration and operation of a vehicle collision time estimation device according to an embodiment of the present invention will be described with reference to the drawings.

[Configuration of vehicle collision time estimation device]
A vehicle collision time estimation device 1 according to an embodiment of the present invention is mounted on a vehicle and, as shown in FIG. 1, includes an imaging unit 2 that captures an image around the vehicle, and an imaging unit 2 that uses a Sobel filter. Edge extraction means 3 for extracting an edge image from captured images that are temporally continuous, edge width normalization means 4 for normalizing the edge width of the edge image extracted by the edge extraction means 3 to a predetermined number of pixels, Voting means 5 for storing information as to how many frames the normalized edge image has been continuously observed as a count-up mask 14 (see FIG. 2), and the moving speed and direction of the edge image using the count-up mask 14 Speed calculation means 6 for calculating the collision time TTC with the object at risk of collision using the position and movement speed of the edge image With the door.

  The imaging means 2 is constituted by an imaging element such as a CMOS or CCD, and is accurate with at least one of the front end portion, rear end portion, and side surface portion of the vehicle, or an object with a risk of collision. It is attached to the location where the collision time should be measured. Further, the imaging unit 2 acquires a surrounding image of the host vehicle at a frame rate equal to or higher than a predetermined value faster than the moving speed of the edge image so that the moving speed and moving direction of the edge image can be easily calculated (details will be described later). To do).

  Further, the edge width normalizing means 4 includes a binarizing means 11, a thinning means 12, and an expanding means 13, as shown in FIG. The moving speed detecting means 6 has a vote value gradient calculating means 15. The functions of the edge extracting means 3, edge width normalizing means 4, voting means 5, moving speed detecting means 6, and collision time calculating means 7 are realized by an in-vehicle computer executing a computer program.

  The vehicle collision time estimation device 1 having such a configuration facilitates calculation of the collision time and obtains a robust output with respect to the position detection error by executing the following collision time calculation process. Enable. Hereinafter, with reference to the flowchart shown in FIG. 3, the operation | movement inside the collision time estimation apparatus 1 for vehicles at the time of performing a collision time calculation process is demonstrated.

[Collision time calculation process]
The flowchart shown in FIG. 3 starts when the imaging unit 2 acquires a vehicle surrounding image for each predetermined frame rate and inputs the acquired vehicle surrounding image to the edge extracting unit 3, and the collision time calculation process is performed in step S1. Proceed to processing.

  In the process of step S1, the edge extraction unit 3 extracts edge images in the horizontal and vertical directions from the vehicle surrounding image captured by the imaging unit 2 using a Sobel filter. Thereby, the process of step S1 is completed, and the calculation process proceeds to the process of step S2.

  In the process of step S2, the binarization means 11 binarizes (0/1) the horizontal and vertical edge images extracted by the process of step S1, as shown in FIG. Apply processing. Thereby, the process of step S2 is completed, and the calculation process proceeds to the process of step S3.

  In the process of step S3, the thinning means 12 thins the active (value 1) edge image to a predetermined pixel width (one pixel in the example shown in FIG. 4B) as shown in FIG. 4B. Turn into. Note that the thinning unit 12 repeatedly performs the thinning process until the edge width reaches a predetermined pixel width. Thereby, the process of step S3 is completed, and the calculation process proceeds to the process of step S4.

  In the process of step S4, the expansion unit 13 expands the edge image to a predetermined pixel width. Specifically, as a result of thinning the edge image, when the edge image is observed at the position x0 as shown in FIG. 4B, the expansion means 13 moves the position x0 as shown in FIG. 4C. By setting the pixels at the positions x0-1 and x0 + 1 on both sides of the image to be active, the edge image is expanded to a predetermined pixel width (3 pixels in the example shown in FIG. 4C). Thereby, the process of step S4 is completed, and the calculation process proceeds to the process of step S5.

  In the processing of steps S5 and S6, the voting means 5 counts up the value (count value) of the memory address corresponding to the position where the standardized edge image is observed (step S5), and no edge is observed. The value of the memory address corresponding to the specified position is reset to 0 (step S6). Specifically, when the edge image shown in FIG. 4C is detected at time t, the voting means 5 detects the positions x0-1 and x0 where the edge image is detected as shown in FIG. , X0 + 1 is incremented by one and the count values at other positions are reset to zero.

  Next, when an edge image is observed at the position x0 even at time t + 1, the voting means 5 counts the positions x0-1, x0, and x0 + 1 at which the edge image is detected, as shown in FIG. Is incremented by one and the voting values at other positions are reset to zero. Next, when the edge image is shifted by one pixel in the positive direction of the x axis and observed at the position x0 + 1 at time t + 2, the voting means 5 detects the position x0 where the edge image is detected as shown in FIG. , X0 + 1, x0 + 2 are incremented by one and the count values at other positions are reset to zero. Thereby, the processing in steps S5 and S6 is completed, and the calculation processing proceeds to processing in step S7.

Here, when the frame rate is sufficiently higher than the moving speed of the edge image, the edge image always has an overlapping area between consecutive frames (in the example shown in FIG. 5, it overlaps with a width of 2 pixels). . Therefore, by counting up the count value at the position where the edge image is observed as described above, the count value becomes equivalent to the time during which the edge image is observed at the same position. When the edge image moves, the count value of the position where the edge image is newly observed is 1, which is the smallest value among the voting values of the edge image. That is, the count value in the direction in which the edge image moves is small, and conversely, the count value in the edge image in the direction opposite to the direction in which the edge image moves is large. Therefore, the gradient generated by the difference between the count values, that is, the gradient α of the straight line Y shown in FIG. 5, indicates how many frames the edge image was observed at the same position until the edge image moved. This value is equivalent to the reciprocal of the moving speed vedge of the edge image as shown in Equation 1 below.

  Specifically, in the example shown in FIG. 5B, if the count values at the positions x0-1, x0, and x0 + 1 are 6, 4, and 2, respectively, the position of the edge image is the position x0 every time it is shifted. It can be seen that H = 6-2 = 4 frames, which are the difference between the count value of −1 and the count value of the position x0 + 1, were continuously observed. Since the vote value h = 2 at the position x0 + 1 after the shift to the position x0, it can be seen that the edge image is continuously observed for two frames. Therefore, it can be seen that the focused edge image moves by one pixel in four frames, and the moving speed of the edge image can be detected. Further, when the frame rate is sufficiently high, it can be assumed that the edge image is moving at a constant speed. Therefore, in the example shown in FIG. 5, the edge image moves by one pixel every four frames, and two frames at the time t + 1. From the continuous observation, it can be seen that the edge image is shifted from the position x0 by 2 frames / {4 frames / 1 pixel} = 0.5 pixels at the time t + 1.

  In the process of step S7, the vote value gradient calculating means 15 calculates the reciprocal of the moving speed vedge of the edge image by calculating the gradient α of the count value. Thereby, the process of step S7 is completed, and the calculation process proceeds to the process of step S8.

In the process of step S8, the collision time calculation means 7 calculates the collision time with the object at risk of collision based on the position of the edge image and its moving speed. Specifically, as shown in FIGS. 6 and 7, the positions of the edge images detected at time t and time t + dt are respectively y, y + dy, and the distance traveled by the vehicle A between time t and time t + dt is dL. In this case, the distance Z from the own vehicle A to the other vehicle B at the time t + dt can be expressed as Equation 2 below. Further, the moving speed vedge in the y direction (vertical direction) of the edge image and the relative speed Vvehicle of the other vehicle B with the host vehicle A can be expressed as Equations 3 and 4 below.

Therefore, the distance Z from the own vehicle A to the other vehicle B at time t + dt is expressed as Equation 5 shown below by substituting Equations 3 and 4 into Equation 2; By substituting the position y of the edge image at the time t and the moving speed vedge (or the gradient α of the count value), the collision time TTC with the other vehicle B can be calculated. That is, by calculating the time when the distance to the other vehicle B is zero without calculating the exact distance to the other vehicle B and the relative speed with the other vehicle B, The collision time TTC is calculated. Thereby, the process of this step S8 is completed and a series of calculation processes are completed.

  According to such a calculation process, since the collision time can be calculated for each point of the edge image, the collision time calculation unit 7 calculates the collision time for each point, and then calculates the collision time according to the collision time. As shown in FIG. 5, the target edge Eobj and the background edge Ebck can be separated from the image. Specifically, in general, the collision time calculated for the background edge Ebck in the distance increases, while the collision time calculated for the target edge Eobj approaching the host vehicle decreases. Therefore, the collision time calculation means 7 can classify the edge associated with the object according to the magnitude of the collision time. In addition, since the collision time becomes large for the object that translates at the same speed as the own vehicle or moves away from the own vehicle, the collision time calculating means 7 does not need to classify the object accurately.

  As is clear from the above description, in the vehicle collision time estimation apparatus 1 according to an embodiment of the present invention, the image pickup means 2 picks up an image around the vehicle, and the edge extraction means 3 picks up an image picked up by the image pickup means 2. The edge image is extracted from the edge image, the edge width normalizing means 4 normalizes the edge width of the edge image extracted by the edge extracting means 3, and the voting means 5 uses the edge width normalized by the edge width normalizing means 4. For the image, the count value corresponding to the position where the edge image is detected is incremented, the count value corresponding to the position where the edge image is not detected is initialized, and the moving speed detecting means 6 is based on the inclination of the count value. The moving direction and moving speed of the edge image extracted by the edge extracting means 3 are calculated, and the collision time calculating means 7 calculates the position of the edge image calculated by the moving speed detecting means 6. Using fine movement speed, calculates the collision time with the object. According to such a configuration, it is possible to calculate the moving speed and the collision time of the edge image without performing a block matching process such as template matching, thereby facilitating the calculation of the collision time and detecting the position. An output that is robust against errors can be obtained.

  Moreover, in the vehicle collision time estimation apparatus 1 according to an embodiment of the present invention, the imaging means 2 is accurate with at least one of the front end portion, the rear end portion, and the side surface portion of the vehicle, or with an object. Since it is attached to the location where the collision time is to be measured, it is possible to eliminate the influence of the position of the photographing means 2 with respect to the collision time calculated when the imaging means 2 is attached offset from both ends of the vehicle.

  Furthermore, in the vehicle collision time estimation device 1 according to an embodiment of the present invention, the collision time calculation means 7 can classify objects in the image based on the calculated collision time, so that there is a collision risk. It is possible to remove a background object detected at substantially the same position as an object.

  As mentioned above, although the embodiment to which the invention made by the present inventors was applied has been described, the present invention is not limited by the description and the drawings that form part of the disclosure of the present invention according to this embodiment. For example, in the above embodiment, the edge width is standardized by performing the thinning process and the expansion process. However, the binary image in which the edge peak position of the edge image is detected and the edge peak position has a predetermined number of pixels. The edge width may be normalized by generating. As described above, it is a matter of course that all other embodiments, examples, operation techniques, and the like made by those skilled in the art based on the above embodiments are included in the scope of the present invention.

It is a block diagram which shows the structure of the collision time estimation apparatus for vehicles used as embodiment of this invention. It is a block diagram which shows the internal structure of the edge width normalization means, voting means, and movement speed detection means which are shown in FIG. It is a flowchart figure which shows the flow of the collision time calculation process used as embodiment of this invention. It is a figure for demonstrating the normalization process by the edge width normalization means shown in FIG. It is a figure for demonstrating the count process by the voting means shown in FIG. It is a figure for demonstrating the collision time calculation process by the collision time calculation means shown in FIG. It is a figure for demonstrating the collision time calculation process by the collision time calculation means shown in FIG. It is a figure for demonstrating the target object classification | category process by the collision time calculation means shown in FIG.

Explanation of symbols

1: vehicle collision time estimation device 2: imaging means 3: edge extraction means 4: edge width normalization means 5: voting means 6: moving speed detection means 7: collision time calculation means 11: binarization means 12: thinning Means 13: Expansion means 14: Count-up mask 15: Vote value gradient calculation means

Claims (4)

A vehicle collision time estimation device that is mounted on a vehicle and estimates a collision time with an object,
An imaging means for capturing an image around the vehicle;
Edge extraction means for extracting an edge image from an image captured by the imaging means;
Edge width normalization means for normalizing the edge width of the edge image extracted by the edge extraction means;
For the edge image normalized by the edge width normalizing means, the count value corresponding to the position where the edge image is detected is incremented, and the count value corresponding to the position where the edge image is not detected is initialized. Voting means,
A moving speed detecting means for calculating the moving direction and moving speed of the edge image extracted by the edge extracting means based on the slope of the count value;
A collision time estimating device for a vehicle, comprising: a collision time calculating means for calculating a collision time with an object using the position and moving speed of the edge image calculated by the moving speed detecting means. The vehicle collision time estimation device according to claim 1,
The imaging means, the front end of the vehicle, the rear end portion, and the vehicle collision time estimation apparatus, characterized in that attached to at least one portion in the side surface portion. The vehicle collision time estimation device according to claim 1 or 2, wherein
The collision time estimation device for a vehicle, wherein the collision time calculation means classifies an object based on the calculated collision time. A vehicle collision time estimation method for estimating a collision time with an object mounted on a vehicle,
Capturing an image around the vehicle;
Extracting an edge image from the captured image;
Normalizing the edge width of the extracted edge image;
For the normalized edge image, incrementing the count value corresponding to the position where the edge image was detected, and initializing the count value corresponding to the position where the edge image was not detected;
Calculating the moving direction and moving speed of the extracted edge image based on the slope of the count value;
A method for estimating a collision time for a vehicle, comprising: calculating a collision time with an object using the calculated position and moving speed of an edge image.

Images