JP6772059B2 - Electronic control devices, electronic control systems and electronic control methods - Google Patents

Description

The present invention relates to electronic control devices, electronic control systems and electronic control methods.

Conventionally, there is a pedestrian detection device that detects a pedestrian from an image captured by an image pickup device such as an in-vehicle camera and calculates the distance to the pedestrian (see, for example, Patent Document 1).

Japanese Unexamined Patent Publication No. 2001-351193

However, in the above-mentioned conventional technique, the position of the pedestrian specified from the captured image may not be accurate, and in such a case, there is a problem that the accurate distance to the pedestrian cannot be calculated.

The present invention has been made in view of the above, and an object of the present invention is to provide an electronic control device, an electronic control system, and an electronic control method capable of improving the calculation accuracy of a distance to a pedestrian.

The present invention includes a detection unit, a search unit, a correction unit, and a calculation unit in an electronic control device. The detection unit detects a person area in which the shape of the person exists from the captured image. The search unit searches for the boundary between the bright region and the dark region in the captured image. The correction unit corrects the lower end position of the person area based on the boundary searched by the search unit. The calculation unit calculates the distance to the person based on the lower end position corrected by the correction unit.

According to the present invention, the accuracy of calculating the distance to a pedestrian can be improved.

FIG. 1 is a diagram showing an outline of a control method performed by an electronic control device. FIG. 2 is a block diagram of an electronic control system. FIG. 3 is a diagram showing an example of the search range. FIG. 4A is a diagram (No. 1) for explaining the abnormality determination process by the binarization unit. FIG. 4B is a diagram (No. 2) for explaining the abnormality determination process by the binarization unit. FIG. 5A is a diagram (No. 1) for explaining the abnormality determination process by the correction unit. FIG. 5B is a diagram (No. 2) for explaining the abnormality determination process by the correction unit. FIG. 6 is a flowchart showing a processing procedure executed by the electronic control device.

The electronic control device, the electronic control system, and the electronic control method according to the embodiment will be described in detail with reference to the accompanying drawings. The present invention is not limited to this embodiment.

In the following, the case of calculating the distance from the own vehicle to the person (hereinafter referred to as pedestrian H) reflected in the captured image based on the captured image input from the camera provided in the own vehicle and capturing the front of the own vehicle. explain.

First, the outline of the control method performed by the electronic control device according to the embodiment will be described with reference to FIG. FIG. 1 is a diagram showing an outline of a control method performed by an electronic control device. The control method is executed by the electronic control device 1 described later in FIG.

As shown in FIG. 1, in the electronic control device 1 according to the embodiment, when the captured image F in which the pedestrian H is captured is input from the camera, first, the person region M in which the shape of the pedestrian H exists from the captured image F Is detected.

For example, the electronic control device 1 detects the person region M from the captured image F by using the person dictionary information that serves as a reference for determining whether or not the object reflected in the captured image F is a pedestrian H.

Then, the electronic control device 1 uses the lower end position Mb of the person region M in the captured image F as the ground contact position Hb of the pedestrian H, and calculates the distance from the own vehicle to the pedestrian H.

At this time, the electronic control device 1 can calculate an accurate distance to the pedestrian H if the ground contact position Hb of the pedestrian in the captured image F and the lower end position Mb in the person area M match. ..

However, as shown in FIG. 1, the lower end position Mb in the person area M and the ground contact position Hb of the pedestrian H may be misaligned. In such a case, the conventional electronic control device cannot correctly calculate the distance from the own vehicle to the pedestrian H.

Therefore, the electronic control device 1 according to the embodiment corrects the lower end position Mb in the person area M after detecting the person area M from the captured image F, and walks using the corrected lower end position Mb of the person area M. It was decided to calculate the distance to person H.

Specifically, as shown in FIG. 1, in the captured image F, the brightness, saturation, brightness and the like are different between the pedestrian H and the region other than the pedestrian H (for example, the road surface). Therefore, in the captured image F, there is a boundary between the bright region and the dark region between the pedestrian H and the other region.

FIG. 1 shows a case where the pedestrian H is the dark region Zd and the other region (road surface) is the bright region Zl. The bright region Zl and the dark region Zd may be reversed between the pedestrian H and other regions depending on the clothes of the pedestrian H, the color of the road surface, and the like.

Therefore, the electronic control device 1 searches for the boundary between the bright region Zl and the dark region Zd from the captured image F, and corrects the searched boundary to be the lower end position Mb. Therefore, the electronic control device 1 can accurately match the lower end position Mb of the person area M with the ground contact position Hb of the pedestrian H.

As a result, the electronic control device 1 can improve the calculation accuracy of the distance to the pedestrian H by calculating the distance to the pedestrian H based on the corrected lower end position Mb.

By the way, in the above-mentioned example, the case where the electronic control device 1 simply searches the boundary between the bright region and the dark region from the captured image F has been described, but the above boundary is obtained from the binarized image obtained by binarizing the captured image F. You can also search for.

Further, in such a case, the electronic control device 1 can also perform the abnormality determination process using the binarization threshold value used for the binarization process. Details of this point will be described later with reference to FIGS. 4A and 4B.

Further, the electronic control device 1 can also perform the abnormality determination process based on the history of the corrected lower end position Mb. Details of this point will be described later with reference to FIGS. 5A and 5B.

Next, the configuration of the electronic control system 50 according to the embodiment will be described with reference to FIG. FIG. 2 is a block diagram of the electronic control system 50. The electronic control system 50 includes an electronic control device 1, a camera 5, and a vehicle control device 40. Further, the electronic control device 1 is connected to the camera 5 and the vehicle control device 40.

The camera 5 includes, for example, an image sensor such as a CCD (Charge Coupled Device) or a CMOS (Complementary Metal Oxide Semiconductor), and for example, images the front of the own vehicle.

The electronic control device 1 includes a control unit 2 and a storage unit 3. The control unit 2 includes a conversion unit 21, a detection unit 22, a search unit 23, a correction unit 24, and a calculation unit 25.

The control unit 2 includes, for example, a computer having a CPU (Central Processing Unit), a ROM (Read Only Memory), a RAM (Random Access Memory), an HDD (Hard Disk Drive), an input / output port, and various circuits.

The CPU of the computer functions as a conversion unit 21, a detection unit 22, a search unit 23, a correction unit 24, and a calculation unit 25 of the control unit 2, for example, by reading and executing a program stored in the ROM.

Further, at least one or all of the conversion unit 21, the detection unit 22, the search unit 23, the correction unit 24, and the calculation unit 25 of the control unit 2 are ASIC (Application Specific Integrated Circuit), FPGA (Field Programmable Gate Array), or the like. It can also be configured with the hardware of.

Further, the storage unit 3 corresponds to, for example, a RAM or an HDD. The RAM or HDD can store the person dictionary information 31, the threshold history information 32, the boundary history information 33, and various programs.

The electronic control device 1 may acquire the above-mentioned program and various information via another computer or a portable recording medium connected by a wired or wireless network.

The person dictionary information 31 is information that serves as a reference for determining whether or not the object reflected in the captured image F is a pedestrian H, and is created in advance by machine learning. The person dictionary information 31 prepares, for example, a predetermined number (for example, several hundred to several 10,000 images) of images of a plurality of types of persons whose shapes are known and images of objects other than the person as learning data.

Subsequently, for example, a HOG (Histogram of Gradient) feature amount is extracted from each prepared image. Then, the above-mentioned prepared image is plotted in a two-dimensional plane based on the extracted HOG features.

Subsequently, for example, a classifier such as an SVM (Support Vector Machine) is used to generate a separation line that separates an image of a person and an image of an object other than the person on a two-dimensional plane. The information of the coordinate axes of the two-dimensional plane and the separation line generated by the classifier becomes the person dictionary information 31.

The feature amount extracted from the image prepared in advance is not limited to the HOG feature amount, and may be a SIFT (Scale Invariant Feature Transform) feature amount. Further, the discriminator used for separating the image of a person and the image of an object other than the person is not limited to SVM, and may be, for example, a discriminator such as AdaBoost. Since the threshold value history information 32 and the boundary history information 33 will be described later with reference to FIGS. 4A and 5A, the description thereof is omitted here.

The conversion unit 21 of the control unit 2 acquires the captured image F from the camera 5 and converts the captured image F into a grayscale image by grayscale the captured image F. Then, the conversion unit 21 outputs the grayscale image to the detection unit 22.

Here, the grayscale conversion is a process of converting each pixel in the captured image F so as to be expressed in each gradation from white to black (for example, 256 gradations) according to the brightness. If the captured image F is not a color image, grayscale conversion is omitted.

The detection unit 22 uses the person dictionary information 31 to detect a person, that is, a person area M (see FIG. 1) in which a pedestrian H exists in the grayscale image input from the conversion unit 21.

For example, when the person area M is detected, the detection unit 22 generates the person area information regarding the person area M and outputs the information to the search unit 23 together with the grayscale image. Further, the detection unit 22 performs a process of outputting the person area information to the correction unit 24.

The person area information is information indicating the position and size of the person area M in the captured image F, and is, for example, an XY coordinate value in which the person area M in the captured image F converted into the XY orthogonal coordinate system exists.

The search unit 23 searches for the boundary between the bright region and the dark region from the captured image F, and outputs the searched boundary to the correction unit 24. Further, the search unit 23 includes a setting unit 23a, a binarization unit 23b, and a determination unit 23c.

The setting unit 23a sets the search range R in the captured image F based on the person area M detected by the detection unit 22. For example, the setting unit 23a associates the XY coordinate values in the captured image F of the set search range R with the grayscale image and outputs them to the binarization unit 23b. A specific example of the processing by the setting unit 23a will be described later with reference to FIG.

The binarization unit 23b binarizes the grayscale image using the binarization threshold Th according to the brightness distribution of each pixel in the search range R set by the setting unit 23a.

First, the binarization unit 23b calculates the binarization threshold Th by using, for example, a discriminant analysis method (so-called “Otsu binarization” method). Specifically, the binarization unit 23b creates a histogram for the distribution of the luminance values of the pixels in the search range R, and calculates the value at which the degree of separation of the luminance values is the largest in the histogram as the binarization threshold Th. ..

As described above, the electronic control device 1 can derive the binarization threshold Th to the optimum value according to the surrounding situation by using the setting determination method for the search range R. As a result, the regions other than the pedestrian H and the pedestrian H can be accurately separated.

Here, the binarization unit 23b stores the binarization threshold Th used for binarization as the threshold history information 32 together with the frame number of the grayscale image. Then, the binarization unit 23b can also determine whether or not to perform the binarization process based on the threshold history information 32. Details of this point will be described later with reference to FIGS. 4A and 4B.

When the binarization unit 23b determines that the binarization process is performed in the above determination, the binarization unit 23b generates a binarized image obtained by binarizing the grayscale image using the binarization threshold Th, and sends the binarization unit 23c to the determination unit 23c. Output. The binarization unit 23b generates a binarized image by, for example, assigning "0" to the pixels in the bright region and "1" to the pixels in the dark region in the captured image F.

The determination unit 23c searches the search range R set by the setting unit 23a in the binarized image input from the binarization unit 23b, and searches for a bright area (area where the pixel is "0") and a dark area (pixels are). The boundary of "1" area) is determined.

For example, the determination unit 23c searches the boundary from the lower end to the upper end of the search range R. Then, when the determination unit 23c detects the boundary, the determination unit 23c generates boundary information regarding the boundary and outputs the boundary information to the correction unit 24. The boundary information is, for example, the XY coordinate value of the boundary in the captured image F.

For example, when the determination unit 23c searches the search range R and detects the boundary between the bright region and the dark region for one pixel, the determination unit 23c generates boundary information of the boundary. That is, the electronic control device 1 can suppress oversight of the boundary by the determination unit 23c by setting the boundary detection condition to one pixel, which is the minimum unit.

However, the detection conditions are not limited to this. For example, the boundary detection condition may be that the boundary between the bright region and the dark region is detected over a plurality of adjacent pixels (for example, three or more pixels). By doing so, it is possible to suppress the determination unit 23c from determining a noise component other than the grounding position Hb as a boundary.

The correction unit 24 corrects the lower end position Mb of the person area M input from the detection unit 22 based on the boundary information generated by the determination unit 23c. For example, the correction unit 24 compares the XY coordinate value of the boundary information input from the determination unit 23c with the XY coordinate value of the lower end position Mb of the person area M detected by the detection unit 22.

Then, when the XY coordinate values of both are different values, the correction unit 24 corrects the XY coordinate value of the lower end position Mb to the XY coordinate value of the boundary information, and returns the corrected XY coordinate value to the calculation unit 25. Output. As a result, the calculation unit 25 can calculate an accurate distance to the pedestrian H.

Here, the correction unit 24 can store the corrected XY coordinate value of the lower end position Mb in the storage unit 3 as the boundary history information 33, and can determine the certainty of the correction based on the boundary history information 33. .. Details of this point will be described later with reference to FIGS. 5A and 5B.

The calculation unit 25 calculates the distance to the pedestrian H based on the corrected XY coordinate values input from the correction unit 24. For example, the calculation unit 25 stores information (mounting height, mounting angle, etc.) regarding the mounting position of the camera 5 as set values in advance, and uses triangulation from the set values and the corrected XY coordinate values described above. The distance from the own vehicle to the pedestrian H is calculated.

Then, the calculation unit 25 outputs the calculated distance to the vehicle control device 40. The vehicle control device 40 controls its own vehicle based on the distance input from the calculation unit 25.

When the distance input from the calculation unit 25 is not more than a predetermined distance (for example, 1 meter or less), the vehicle control device 40 controls the brake of the own vehicle to stop the own vehicle.

As a result, it is possible to prevent a collision between the own vehicle and the pedestrian H. The predetermined distance is not limited to the above example, and can be set to any value. Further, the vehicle control device 40 may control the steering of the own vehicle so as to avoid a collision with the pedestrian H, for example.

Next, the search range R set by the setting unit 23a will be described with reference to FIG. FIG. 3 is a diagram showing an example of the search range R.

As shown in the figure, the setting unit 23a sets the vicinity of the lower end position Mb of the person area M as the search range R. Specifically, the setting unit 23a sets, for example, a width of ± 20% of the size of the person area M as the search range R centered on the lower end position Mb.

That is, the setting unit 23a sets the search range R near the lower end of the person area M. As a result, the determination unit 23c intensively searches the area near the lower end of the person area M.

Therefore, the determination unit 23c can efficiently determine the boundary. Further, by limiting the search range R in advance prior to the boundary search process by the determination unit 23c, the processing load on the search of the determination unit 23c can be reduced.

The search range R described above is only an example, and the setting unit 23a may set the search range R to be larger or smaller.

Next, the abnormality determination process by the binarization unit 23b will be described with reference to FIGS. 4A and 4B. 4A and 4B are diagrams for explaining the abnormality determination process by the binarization unit 23b.

Note that FIG. 4A is an example of the threshold history information 32, in which the vertical axis shows the value of the binarization threshold Th, and the horizontal axis shows the frame number corresponding to the horizontal axis. Further, FIG. 4B shows an example of the grayscale image at the frame number F1 and the frame number F2 shown in FIG. 4A.

The binarization unit 23b performs the binarization threshold value as an abnormality determination process when the time-series fluctuation amount of the binarization threshold value Th exceeds a predetermined value ThA based on the threshold value history information 32 shown in FIG. 4A. Th is determined as abnormal.

In the example of FIG. 4A, the fluctuation amount A1 of the binarization threshold Th exceeds the predetermined value ThA between the frame number F1 and the frame number F2. At this time, the binarization unit 23b determines that the binarization threshold Th of the frame number F2 is abnormal.

This is because, as shown in FIG. 4B, noise is mixed in a region other than the pedestrian H, such as the shadow Cs of another vehicle C or another vehicle C being reflected in the search range R between the frame numbers F1 and the frame number F2. This is because

In such a case, the boundary between the bright region and the dark region may exist on the lower end side of the pedestrian H. Therefore, the determination unit 23c may determine the boundary other than the lower end position Mb of the pedestrian H.

Therefore, the binarization unit 23b does not execute the binarization process for the grayscale image of the frame number F2 as the abnormality handling process, and the lower end position of the person area M detected by the detection unit 22 with respect to the correction unit 24. It is instructed to output the XY coordinate value of Mb to the calculation unit 25.

As a result, it is possible to prevent the determination unit 23c from erroneously determining the boundary. Therefore, it is possible to prevent erroneous correction by the correction unit 24. Further, in such a case, the binarization process by the binarization unit 23b and the process by the determination unit 23c can be omitted.

That is, the processing load by the binarization unit 23b and the determination unit 23c can be suppressed while ensuring the accuracy of the correction by the correction unit 24. In this way, the binarization unit 23b not only appropriately separates the pedestrian H from the other region by using the discriminant analysis method, but also performs the abnormality determination process based on the history of the binarization threshold Th. You can also do it.

The frame number F1 and the frame number F2 described above may be continuous or may be separated by an arbitrary number of frames. This also applies to the frame numbers F3 and F4, which will be described later with reference to FIGS. 5A and 5B.

The binarization unit 23b narrows the search range R set by the setting unit 23a to set the binarization threshold Th as an abnormality determination process, and makes the above determination for the binarization threshold Th. You may. As a result, noise is less likely to be mixed in the search range R, so that the binarization unit 23b can set the binarization threshold Th of a normal value.

Next, the abnormality determination process by the correction unit 24 will be described with reference to FIGS. 5A and 5B. 5A and 5B are diagrams for explaining the abnormality determination process by the correction unit 24.

FIG. 5A is an example of the boundary history information 33 shown in FIG. In FIG. 5A, the vertical axis shows the value of the Y coordinate of the corrected boundary, and the horizontal axis shows the frame number corresponding to the horizontal axis. Further, FIG. 5B shows an example of the binarized image at the frame number F3 and the frame number F4 shown in FIG. 5A.

As an abnormality determination process, the correction unit 24 determines the lower end position Mb when the time-series fluctuation amount of the Y coordinate of the lower end position Mb exceeds a predetermined value ThD based on the boundary history information 33 shown in FIG. 5A. Judge as abnormal.

In the example of FIG. 5A, the fluctuation amount D1 of the Y coordinate value exceeds the predetermined value ThD between the frame number F3 and the frame number F4. At this time, the correction unit 24 determines that the lower end position Mb of the frame number F4 is abnormal.

This is because, as shown in FIG. 5B, it is highly possible that noise that was not in the search range R in frame number F3 was generated in frame number F4, and the determination unit 23c determined the noise in frame number F4 as a boundary. Is. That is, when the amount of time-series fluctuation of the value of the Y coordinate is large, the reliability of the boundary determined by the determination unit 23c decreases.

Therefore, in such a case, the correction unit 24 excludes the corrected lower end position Mb from the correction result candidates as an abnormality handling process. Then, for example, the correction unit 24 outputs the XY coordinate value of the lower end position Mb of the person area M detected by the detection unit 22 as a correction value to the calculation unit 25 instead of the boundary information input from the determination unit 23c. ..

Therefore, it is possible to avoid the unreliable correction by the correction unit 24, and it is possible to improve the reliability of the correction. The correction unit 24 may use the coordinates closer to the previous coordinates among the coordinates of the lower end position Mb detected by the detection unit 22 and the boundary determined by the determination unit 23c as the correction result.

Further, although the case where the correction unit 24 determines the abnormality based on the value of the Y coordinate has been described here, the correction unit 24 can also perform the abnormality determination process for the value of the X coordinate.

Further, the abnormality handling process by the correction unit 24 is not limited to the above example. For example, the correction unit 24 may instruct the determination unit 23c to continue the search for the boundary in a direction in which the amount of fluctuation of the lower end position Mb becomes smaller.

Then, when the determination unit 23c newly determines the boundary based on the instruction, the correction unit 24 corrects the lower end position Mb to the boundary.

As a result, the accuracy of correction by the correction unit 24 can be improved. When the determination unit 23c does not newly determine the boundary, the correction unit 24 may use the detection result of the detection unit 22 as the correction result.

Further, for example, the determination unit 23c may search all the boundaries in advance, and then the correction unit 24 may correct the lower end position Mb from all the boundaries to the most probable boundary.

Further, the correction unit 24 derives the XY coordinate values predicted from the boundary history information 33, outputs the XY coordinate values to the setting unit 23a, and instructs the setting unit 23 to set the search range R based on the XY coordinate values. You may decide to do it.

In other words, the setting unit 23a may set the search range R based on the boundary history information 33. As a result, the search range R can be appropriately set in the setting unit 23a, so that the accuracy of processing by the binarization unit 23b and the determination unit 23c can be improved.

In this way, the electronic control device 1 is determined to perform the abnormality determination process by the two processing units of the binarization unit 23b and the correction unit 24. In other words, the electronic control device 1 performs the abnormality determination process before and after the boundary search process by the determination unit 23c.

Therefore, it is possible to improve the accuracy of the boundary determination process by the determination unit 23c and avoid erroneous correction by the correction unit 24. The abnormality determination process may be performed only by one of the binarization section 23b and the correction section 24. In such a case, the processing load of the control unit 2 can be suppressed.

Next, a processing procedure executed by the electronic control device 1 according to the embodiment will be described with reference to FIG. FIG. 6 is a flowchart showing a processing procedure executed by the electronic control device 1 according to the embodiment. It should be noted that such processing is repeatedly executed by the control unit 2 of the electronic control device 1.

As shown in the figure, first, the conversion unit 21 acquires the captured image F from the camera 5 and grayscales the captured image F (step S101).

Subsequently, the detection unit 22 detects the person region M from the grayscale image (step S102). After that, the setting unit 23a sets the search range R based on the person area M (step S103), and the binarization unit 23b calculates the binarization threshold Th (step S104).

Then, the binarization unit 23b determines whether or not the binarization threshold Th calculated as shown in FIGS. 4A and 4B is normal (step S105). Here, the binarization unit 23b binarizes the grayscale image when the binarization threshold Th is normal (step S105, Yes) (step S106).

Subsequently, the determination unit 23c searches the search range R and determines the boundary (step S107). After that, the correction unit 24 determines whether or not the coordinate values of the boundary determined by the determination unit 23c are normal as shown in FIGS. 5A and 5B (step S108).

In such a determination, when the correction unit 24 determines that the coordinate values of the boundary are normal (steps S108, Yes), the correction unit 24 corrects the lower end position Mb of the person area M (step S109). Then, the calculation unit 25 calculates the distance to the pedestrian H (step S110), and ends the process.

On the other hand, in the determination of step S105, when the binarization threshold Th is abnormal (step S105, No), the calculation unit 25 is a pedestrian based on the lower end position Mb of the person area M detected by the detection unit 22. The distance to is calculated (step S110).

Further, in the determination of step S108, if the coordinate value of the boundary is abnormal (steps S108, No), the electronic control device 1 repeats the processes after step S107.

When the coordinate values of the boundary are abnormal (steps S108, No), as described above, the calculation unit 25 reaches the pedestrian H based on the lower end position Mb of the person area M detected by the detection unit 22. It may be decided to calculate the distance of.

As described above, the electronic control device 1 according to the present embodiment includes a detection unit 22, a search unit 23, a correction unit 24, and a calculation unit 25. The detection unit 22 detects a person region M in which the shape of a person (pedestrian H) exists from the captured image F. The search unit 23 searches for the boundary between the bright region and the dark region in the captured image F. The correction unit 24 corrects the lower end position Mb of the person area M based on the boundary searched by the search unit 23. The calculation unit 25 calculates the distance to the person (pedestrian H) based on the lower end position Mb corrected by the correction unit 24. Therefore, according to the electronic control device 1 according to the embodiment, it is possible to improve the calculation accuracy of the distance to the pedestrian H.

By the way, in the above-described embodiment, the case where the binarization unit 23b binarizes each pixel based on the brightness of the captured image has been described, but the captured image is binarized according to the brightness and saturation of the captured image. It may be binarized.

Further, in the above-described embodiment, the case where the search unit 23 determines the boundary between the bright region and the dark region from the binarized image has been described, but the present invention is not limited to this.

For example, the search unit 23 may extract the edge of the pedestrian H from the captured image F and determine it as the boundary between the bright region and the dark region based on the edge. For example, a Sobel filter, a Laplacian filter, or the like can be used as the edge extraction method. In such a case, for example, an infrared camera may be used as the camera 5.

Further, in the above-described embodiment, the case where the camera 5 images the front of the own vehicle has been described, but the distance to the pedestrian H in the captured image input from the camera that images the side or the rear of the own vehicle. It can also be applied to the case of calculating. Further, in the present embodiment, it is also possible to calculate the distance to the pedestrian H from the captured image input from the fixed camera provided in the street light, the building, or the like.

Further effects and variations can be easily derived by those skilled in the art. For this reason, the broader aspects of the invention are not limited to the particular details and representative embodiments expressed and described above. Therefore, various changes are possible without departing from the spirit or scope of the overall concept of the invention as defined by the appended claims and their equivalents.

1 Electronic control device 5 Camera 21 Conversion unit 22 Detection unit 23 Search unit 23a Setting unit 23b Binarization unit 23c Determination unit 24 Correction unit 25 Calculation unit 40 Vehicle control system 50 Electronic control system M Person area Mb Lower end position H Pedestrian ( An example of a person)
Hb grounding position

Claims (7)

A detection unit that detects the human area where the shape of the person exists from the captured image,
A search unit for searching the boundary between the bright region and the dark region in the captured image,
A correction unit that corrects the lower end position of the person area based on the boundary searched by the search unit, and a correction unit.
An electronic control device including a calculation unit that calculates a distance to the person based on the lower end position corrected by the correction unit. The correction unit
The first aspect of the present invention is to exclude the corrected lower end position from the candidates for the correction result when the time-series fluctuation amount of the lower end position after the correction using the boundary exceeds a predetermined value. The electronic control device described. The search unit
The electronic control device according to claim 2, wherein when the time-series fluctuation amount of the lower end position exceeds a predetermined value, the search for the boundary is continued in the direction in which the fluctuation amount becomes smaller. The search unit
A setting unit that sets a search range based on the person area detected by the detection unit, and a setting unit.
A binarization unit that binarizes the captured image using a binarization threshold value determined based on the brightness distribution of each pixel in the search range set by the setting unit.
The electronic control according to claim 1, 2 or 3, further comprising a determination unit that searches the search range of the captured image binarized by the binarization unit and determines the boundary. apparatus. The binarization unit
The claim is characterized in that when the time-series fluctuation amount of the binarization threshold value exceeds a predetermined value, the correction unit is instructed to use the lower end position detected by the detection unit as the correction result. The electronic control device according to 4. The electronic control device according to any one of claims 1 to 5 provided on the vehicle, and
An image pickup device that captures the captured image around the vehicle, and
An electronic control system including a vehicle control device that controls a brake of the vehicle based on the distance calculated by the electronic control device. A detection process that detects a human area in which the shape of a person exists from an captured image,
A search step for searching the boundary between the bright region and the dark region in the captured image, and
A correction step of correcting the lower end position of the person area based on the boundary searched by the search step, and a correction step of correcting the lower end position of the person area.
An electronic control method including a calculation step of calculating a distance to the person based on the lower end position corrected by the correction step.

Images