JP7192312B2 - Image processing device - Google Patents

Description

The present disclosure relates to an image processing device.

Patent Literature 1 describes a technique for recognizing targets such as pedestrians by pattern matching. According to the pattern matching, an area of a predetermined size is extracted from an image, and the target area, which is the extracted area, is divided into grid-like cells. Then, the HOG feature quantity, which is a feature quantity obtained by histogramming the direction of the luminance gradient, is calculated for each cell divided into a lattice. Further, the HOG feature amount for each cell is compared with the HOG feature amount of a matching pattern, which is a pattern corresponding to the part of the target to be recognized and stored in advance. If the HOG feature values match as a result of matching, the cell is determined to represent the part of the target represented by the matching pattern. Then, it is determined that there is a target including a portion of the target corresponding to the collation pattern in the range including the cell in the image.

JP 2014-81826 A

However, for example, when pattern matching is performed using the above-mentioned HOG feature quantity for each of a target region including a human leg and a target region including a sign post, both the human leg and the sign post are relative to the ground. Since there is a luminance gradient in the horizontal direction, the HOG feature quantities obtained by histogramming the direction of the luminance gradient are similar.

As a result, pattern matching using HOG features cannot distinguish between a target region including a person's leg and a target region including a sign post. is recognized as a target area containing a

One aspect of the present disclosure provides a technique for improving target recognition accuracy.

One aspect of the present disclosure is an image processing apparatus comprising an image acquisition unit (S210), a range setting unit (S110, S310), an edge detection unit (S340), a distribution calculation unit (S370), a distribution pattern A memory (64) and a target determination unit (S380) are provided. The image acquisition unit is configured to acquire a captured image (100) captured by an imaging device (10) mounted on a vehicle. The range setting section is configured to set, as a target range, a range in which the target is estimated to exist in the captured image acquired by the image acquisition section. The edge detection unit is configured to detect, in at least part of the target range set by the range setting unit, as edge points representing the outline portion of the target along the horizontal direction in the image. . The distribution calculation unit calculates the edge degree, which is a numerical value obtained by adding the number of edge points calculated by the edge detection unit in the vertical direction of the image, to an edge, which is a distribution arranged along the horizontal direction of the image. configured to calculate a distribution; The distribution pattern memory is configured to store at least one distribution pattern, which is a representative pattern of edge distribution pre-associated with each type of target. The target determination unit determines a target range corresponding to the distribution pattern according to the degree of similarity between the edge distribution calculated by the distribution calculation unit and each of at least one distribution pattern stored in the distribution pattern memory. It is configured to determine whether it is what it represents.

In pattern matching using HOG features, the target area to be matched is divided into a plurality of cells, and the matching results are totaled for each divided cell, so that the entire target area matches the target. Determine whether or not Therefore, if a portion of each divided cell unit has a high degree of matching with the target to be verified, the result of the cell with a high degree of matching affects the result of the entire target area. may be determined to match the

According to the configuration described above, the distribution calculation unit calculates the edge distribution, which is a distribution in which edge degrees, which are numerical values obtained by adding in the vertical direction in the image, are arranged along the horizontal direction in the image. As a result, the edge distribution in a wider range is calculated, at least in the vertical direction, compared to the case where the determination is made in units of cells. recognize.

Since comparison is possible based on detection results in a wider range than in the case of making determinations on a cell-by-cell basis, even if the degree of matching is high for a part corresponding to a cell, the effect on the whole can be suppressed.
As a result, it is possible to improve the accuracy of recognizing the target represented by the target range.

It should be noted that the symbols in parentheses described in this column and the scope of claims indicate the correspondence with specific means described in the embodiment described later as one mode, and the technical scope of the present disclosure is It is not limited.

1 is a block diagram showing the configuration of an image recognition system; FIG. It is a flow chart of pedestrian recognition processing. 6 is a flowchart of estimation processing; 4 is a flowchart of recognition processing; FIG. 4 is a diagram showing an example of an estimated range and an attention range in a captured video and a captured image; FIG. 4 is a diagram showing an image of a first attention range; FIG. 4 is a diagram showing an edge image of a first attention range; FIG. 10 is a diagram showing edge distribution of the first attention range; FIG. 11 is a diagram showing an image of a third attention range; FIG. 11 is a diagram showing an edge image of a third attention range; FIG. 11 is a diagram showing edge distribution of a third attention range;

Hereinafter, embodiments of the present disclosure will be described with reference to the drawings.
[1. composition]
An image recognition system 1 shown in FIG. 1 is mounted on a vehicle and includes a camera module 10, a control device 20, a brake actuator 30, a display device 40, a speaker 50 and an image processing device 60. Hereinafter, the vehicle on which the image recognition system 1 is mounted is referred to as the own vehicle.

The camera module 10 is, for example, a camera module of a CCD camera that outputs color images. The camera module 10 is provided in the interior of the vehicle, for example, to capture an image of the front of the vehicle.

Further, the camera module 10 is not limited to one that captures an image of the front of the vehicle, but may be one that captures an image of the surroundings of the vehicle.

The brake actuator 30 is an actuator that adjusts the braking force of the host vehicle.
The display device 40 is installed in the passenger compartment of the vehicle and displays an image or the like on a display screen of the display device 40 .

The speaker 50 is a device that is installed inside the vehicle and outputs sound.
The control device 20 controls the brake actuator 30 , the display device 40 or the speaker 50 according to the pedestrian recognition result output by the image processing device 60 .

Control of the brake actuator 30 refers to brake control that increases the braking force of the vehicle when it is determined that a pedestrian exists within a predetermined range based on the position of the vehicle.

The control of the display device 40 is, for example, when a captured image of the surroundings of the own vehicle is displayed on the display device 40, and a pedestrian range, which is a range in which it is determined that a pedestrian exists in the captured image, is present. In addition, it refers to control that emphasizes and displays the range. Note that the control of the display device 40 is not limited to the control of performing the highlighted display, and may be the control of performing the display to call attention to the driver of the own vehicle. Further, when there is an object range, which is a range in which it is determined that an object exists in the captured image, the range may be emphasized and displayed. In this case, the pedestrian range display and the object range display may be displayed in different display modes.

The control of the speaker 50 means, for example, when it is determined that a pedestrian exists around the vehicle imaged by the camera module 10, a warning sound or the like is emitted to notify the driver of the vehicle.

Note that the control device 20 may perform only one of the control of the brake actuator 30, the control of the display device 40, and the control of the speaker 50, or may perform a plurality of controls.

The image processing apparatus 60 includes a microcomputer having a CPU 61 and a semiconductor memory such as RAM or ROM (hereinafter referred to as processing memory 62), a collation pattern memory 63, and a distribution pattern memory 64. FIG. Each function of the image processing device 60 is realized by the CPU 61 executing a program stored in a non-transitional substantive recording medium. In this example, the processing memory 62 corresponds to the non-transitional substantive recording medium storing the program. Also, by executing this program, a method corresponding to the program is executed. Note that the image processing device 60 may include one microcomputer, or may include a plurality of microcomputers.

The method of realizing each function provided in the image processing device 60 is not limited to software, and some or all of the functions may be realized using one or more pieces of hardware. For example, when the above functions are realized by an electronic circuit that is hardware, the electronic circuit may be realized by a digital circuit, an analog circuit, or a combination thereof.

The image processing device 60 is supplied with power from the own vehicle while the ignition switch of the own vehicle is on. The image processing device 60 also performs processing for outputting detection results to the control device 20 . The detection result here includes, for example, whether or not a pedestrian exists, and may include information on the position of the detected pedestrian.

A collation pattern memory 63 is a memory for storing a collation pattern. The matching pattern is a rectangular pattern corresponding to a portion of the pedestrian, and represents a predetermined HOG feature amount. The HOG feature amount is a histogram representing the gradient strength of the luminance gradient for each gradient direction of the luminance gradient.

The distribution pattern memory 64 is a memory for storing distribution patterns. The distribution pattern is a representative pattern of edge distribution of at least a part of the target that is set in advance for each type of target. The edge distribution is a distribution in which edge degrees are arranged along the horizontal direction in the captured image. The edge degree is a value calculated by adding the number of edge points along the vertical direction in the captured image.

That is, the distribution pattern is a representative pattern of edge distribution stored in advance for each type of target in order to be compared with the edge distribution of a portion of the target detected in the captured image.

Furthermore, the distribution pattern includes object patterns. The object pattern is a distribution pattern representing at least part of objects other than pedestrians, which are targets. The object pattern includes, for example, a pole pattern, which is an object pattern representing a portion of a pole, which is a columnar body erected on a road.

Note that the camera module 10 corresponds to an imaging device.
[2. process]
Next, pedestrian recognition processing executed by the CPU 61 will be described with reference to the flowchart of FIG. The pedestrian recognition device is repeatedly executed while the image processing device 60 is powered on.

In S110, the CPU 61 performs estimation processing. Details of the estimation processing here will be described later. It refers to the process of extracting as Note that when a plurality of estimated ranges exist in the captured image, the plurality of estimated ranges existing in the captured image are extracted.

In S120, the CPU 61 determines whether or not an estimated range exists in the captured image by the estimation processing of S110, that is, whether or not one or more estimated ranges are extracted by the estimation processing.
When it is determined in S120 that there is an estimated range, that is, when it is determined that one or more estimated ranges are extracted, the CPU 61 shifts the process to S130.

On the other hand, if the CPU 61 determines in S120 that no estimated range exists, that is, if no estimated range has been extracted, the pedestrian recognition process ends.
In S130, the CPU 61 performs recognition processing. The details of the recognition processing here will be described later, but in the recognition processing, each of the estimated ranges extracted in S110 is a pedestrian range representing pedestrians or a non-pedestrian range. This is the process of determining whether the object range is a real object and outputting the pedestrian range and the object range.

S110 corresponds to the processing of the range setting unit.
<Estimation processing>
Next, details of the estimation process executed by the CPU 61 in S110 of the pedestrian recognition process will be described with reference to FIG.

In S<b>210 , the CPU 61 acquires the captured image captured by the camera module 10 .
In S220, the CPU 61 divides the captured image acquired in S210 into rectangular cells having a preset number of pixels, and calculates the HOG feature amount for each cell. Here, the number of pixels of the cell to be set is set within a range of 16 pixels×16 pixels, for example.

Note that the HOG feature amount is obtained as follows. That is, assuming that the abscissa of the cell is x, the ordinate is y, and the luminance at the coordinates (x, y) is L(x, y), first, the gradient strength m (x, y) and the gradient direction θ(x, y) representing the direction of the luminance gradient of each divided area are calculated.

Then, 0° to 180° are divided in the m direction, and a brightness gradient histogram is created in which the sum of the gradient strengths of the divided regions that can be regarded as having the same gradient direction is the gradient strength in that direction. do. The m-dimensional value represented by this luminance histogram is the HOG feature amount.

In S230, the CPU 61 determines whether or not an unmatched pattern, which is a matching pattern not selected in S240 to be described later, exists in the captured image acquired in S210.
When the CPU 61 determines that there is an unmatched pattern, the process proceeds to S240.

Through the processing from S240 to S270, the area in the captured image is compared with each matching pattern stored in the matching pattern memory 63, and the estimated range representing the pedestrian present in the captured image is extracted.

At S240, the CPU 61 selects one matching pattern from among the unmatched patterns.
In S250, the CPU 61 determines whether or not there is an unselected exploration size (hereinafter referred to as an unselected size) that has not been selected in S260 to be described later among the plurality of exploration sizes set for the exploration area. Here, the search area is a rectangular area represented by n cells×n cells, where n is a natural number. As for the size of the search area, a plurality of types are set according to the size of the target to be searched and reflected in the captured image.

If the CPU 61 determines in S250 that there is no unselected size, it returns to S230 and performs subsequent processing. That is, the process from S260 to S270 is repeatedly executed by changing the collation pattern.

On the other hand, when the CPU 61 determines in S250 that there is an unselected size, the process proceeds to S260.
In S260, the CPU 61 selects any one search size from the unselected sizes. The search size selected in S260 is hereinafter also referred to as the selected size.

In S270, the CPU 61 extracts the estimated range. Specifically, extraction of the estimated range is performed as follows. That is, the entire captured image is scanned while shifting the search area of the selected size by one cell. At this time, the sum of the HOG feature amounts of all the cells included in the search area is used as the HOG feature amount of the search area, and the similarity is calculated by comparing the HOG feature amount of the search area and the HOG feature amount of the matching pattern. calculate. The position of the search area with a high degree of similarity is extracted as an estimated range in which the pedestrian represented by the matching pattern may exist.

When the CPU 61 finishes scanning the entire captured image, the process proceeds to S250. That is, the CPU 61 changes the selected size and repeats the same processing.
If it is determined in S230 that there is no unmatched pattern, the estimation process ends.

It should be noted that S210 corresponds to processing as an image acquisition unit.
<Recognition processing>
Next, the details of the recognition process executed by the CPU 61 in S130 of the pedestrian recognition process will be described with reference to FIG.

At S310, the CPU 61 acquires all estimated ranges extracted at S270 of the estimation process. Here, as described above, the estimated range is the range estimated to be a pedestrian within the target range.

In the processing from S320 to S400, it is determined whether or not the target range, which is at least a part of each estimated range, matches the object pattern. It is determined whether or not represents an object.

In S320, the CPU 61 determines whether or not there is an unprocessed range, which is an estimated range not selected in S330, among all the estimated ranges.
When the CPU 61 determines that there is an unprocessed range, the process proceeds to S330.

In S330, the CPU 61 selects one estimated range from the unprocessed ranges as a selection range.
In S340, the CPU 61 detects, as an edge point, a point where the absolute value of the luminance difference between horizontally adjacent pixels in the range selected in S330 is equal to or greater than a predetermined luminance threshold. . Note that the brightness threshold value is set to a value corresponding to the brightness difference of the contour portion of the target.

At S350, the CPU 61 determines whether or not there is an uncalled pattern that has not been called at S360 to be described later among the distribution patterns.
At S<b>360 , the CPU 61 calls one of the object patterns stored in the distribution pattern memory 64 . The object pattern called up in S360 is called out of the object patterns that have not yet been called up for the estimated range selected in S330.

In S370, the CPU 61 detects each boundary between pixel rows arranged in the vertical direction in the range of interest, which is a predetermined range constituting at least a part of the selection range selected in S330. The number of edge points obtained is calculated as the edge degree. Then, a distribution obtained by arranging the calculated edge degrees in the horizontal direction corresponding to the position of the boundary of the pixel row is derived as an edge distribution.

The attention range is set to a range corresponding to the object pattern called in S360 in the selection range. Here, the range of interest set corresponding to the object pattern is set in advance to a range in which a part of the target indicated by the object pattern is likely to be recognized as a part of the pedestrian.

At S380, the CPU 61 determines whether or not the calling pattern, which is the object pattern called at S360, matches the edge distribution calculated at S370.
Here, the match between the call pattern and the edge distribution is not limited to perfect match, and may be determined as match if the degree of similarity is equal to or greater than a predetermined threshold. The similarity may also be obtained by comparing the distribution characteristics of the call pattern and the distribution characteristics of the edge distribution. Here, the characteristics of the distribution include, for example, the number of peaks, the spread of the distribution, the positions of the peaks, and the like. Here, the peak means, for example, a portion where the degree of edge is the maximum or a portion where the degree of edge is the maximum. For example, the half width of the peak may be used as the spread of the distribution. The similarity of the distributions may be regarded as matching if, for example, the edge distribution and the distribution pattern have the same number of edge degree peaks and the spread of the edge degree distributions is similar.

When the CPU 61 determines in S380 that the edge distribution and the calling pattern match, the process proceeds to S390.
At S390, the CPU 61 recognizes that the selection range selected at S330 is an object range representing an object, and returns to S320 to execute subsequent processing.

On the other hand, when the CPU 61 determines in S380 that the edge distribution and the calling pattern do not match, the process proceeds to S350.
Returning to S350, if it is determined that there is no uncalled pattern, the process proceeds to S400.

At S400, the CPU 61 recognizes that the selection range selected at S330 is a pedestrian range representing a pedestrian, returns to S320, and executes subsequent processing.
In this way, by the processing from S350 to S400, the attention range, which is a part of the selection range selected in S330, and the object pattern, which is a pattern representing a part of the object, are compared to determine whether the selection range represents the object. It is determined whether there is

If it is determined in S320 that there is no unprocessed range, the process proceeds to S410.
At S410, the CPU 61 outputs the pedestrian range and the object range, and terminates the estimation process.

Through the recognition process, the CPU 61 outputs the estimated range other than the object range as the pedestrian range.
S310 corresponds to processing as a range setting unit, S360 corresponds to processing as a range extraction unit, S340 and S370 corresponds to processing as an edge detection unit, and S370 corresponds to processing as a distribution calculation unit. do. Further, S380 corresponds to the processing of the target determination unit, and S400 corresponds to the processing of the pedestrian determination unit.

[3. Operation example]
An example of image processing when the pedestrian recognition processing is performed will be described with reference to the drawings.
As shown in FIG. 5, for example, as a captured image, a stationary pedestrian 101, a second pedestrian 102 walking across the front of the vehicle, and a pole 103 are shown in the image. Take the included captured image 100 as an example. Note that the captured image 100 is not limited to the image shown in FIG. 5, and may be various images captured around the vehicle.

When the estimation process is performed on the captured image 100 in S110 of the pedestrian recognition process, for example, the first estimated range 101a, the second estimated range 102a, and the third estimated range 103a in FIG. 5 are extracted as estimated ranges.

Here, when extracting the estimated range with the HOG feature amount, not only the first pedestrian 101 and the second pedestrian 102 but also the pole 103 may be extracted as the estimated range.
The rod-like shape of the pole included in the attention range, which is part of the estimation range, has a shape whose longitudinal direction is the vertical direction with respect to the image, similar to human legs, and whose luminance is horizontal with respect to the image. have a gradient. For this reason, there is a possibility that the rod-like shape of the pole will be recognized as a human leg.

That is, when pedestrians are detected by comparing HOG feature amounts, not only pedestrians but also three-dimensional objects such as poles installed on the road may be extracted.
Next, in S130, recognition processing is performed for each estimated range. In the recognition process, when the first estimated range 101a is selected in S330, the first estimated range 101a is compared with the object pattern.

Recalling the object pattern is performed at S360. At S340, edge point detection is performed on the selection.
Detection of edge points is done as follows. That is, an edge image as shown in FIG. 7 is created for the selected range as shown in FIG. The term "edge image" as used herein refers to an image representing an edge point, which is a point whose absolute value of luminance difference in the horizontal direction is equal to or greater than a predetermined luminance threshold. A plurality of white points Pe1 in FIG. 7 represent edge points.

By adding the number of edge points along the vertical direction using the edge image shown in FIG. 7, the edge degree for each boundary of the pixel row in the selection range is calculated. By arranging the calculated number of edge points along the horizontal direction of the image for each pixel row boundary, an edge distribution representing the edge degree for each pixel row boundary in the range of interest as shown in FIG. 8 is derived. be done.

At S380, it is determined whether the edge distribution derived at S370 matches the object pattern representing part of the pole, which is the calling pattern. As a feature of Paul's object pattern, for example, if it is stored that the edge distribution has two edge degree peaks, the edge distribution derived in S370 has three or more peaks. It is determined that it does not match the object pattern representing the pole.

As described above, when it is determined that the first attention range 101b, which is a part of the first estimated range 101a, does not match any of the object patterns, it is recognized as a range representing a pedestrian.
As with the first estimated range 101a, the second estimated range 102a is also compared with the object pattern in the second attention range 102b to determine whether or not the range represents a pedestrian.

On the other hand, in the recognition process, when the third estimated range 103a is selected in S330, matching with the object pattern is performed in the third attention range 103b as in the case of the first estimated range 101a.

Here, by the edge point detection in S340, a point Pe2 representing an edge point as shown in FIG. An edge image containing When the edge distribution is obtained using the edge image shown in FIG. 10 in the same manner as in the case of the first attention range 101b, the edge distribution representing the degree of edge for each pixel row boundary is derived as shown in FIG. .

At S380, it is determined whether the edge distribution derived at S370 matches the object pattern representing part of the pole, which is the calling pattern. Similarly, if it is stored that the edge distribution has two edge degree peaks as a feature of Paul's object pattern, the edge distribution derived in S370 has two peaks. It is determined to match the features of the object pattern representing the pole.

As a result, the third estimated range 103a is determined to match the object pattern representing the pole, and is recognized as the object range representing the object.
Then, when it is determined that recognition processing has been performed for all the estimation ranges in the captured image 100, namely the first estimation range 101a, the second estimation range 102a, and the third estimation range 103a, in S410, S first estimation A recognition result indicating that the range 101a and the second estimated range 102a are pedestrian ranges and the third estimated range 103a is an object range is output.

[4. effect]
According to the embodiment detailed above, the following effects are obtained.
(1) According to the above embodiment, an edge distribution that is a distribution of edge degrees representing the position of the outline of the target in the cell and a predetermined distribution that is a distribution of edge degrees representing the position of the outline of the target. Patterns are compared. Therefore, the target is determined using the position of the outline of the target as a determination material, compared to the case where the HOG feature amount of the cell and the matching pattern is compared.

As a result, it is possible to improve the accuracy of recognizing the target represented by the target range.
For example, even targets such as a pedestrian and a pole that cannot be identified by pattern matching using the HOG feature amount can be identified according to the above embodiment.

[5. Other embodiments]
Although the embodiments of the present disclosure have been described above, the present disclosure is not limited to the above-described embodiments, and various modifications can be made.

(1) In the estimation process, whether or not a pedestrian is estimated based on whether or not the HOG feature values match. The feature amount is not limited to the feature amount, and may be any feature amount corresponding to the feature of each cell representing the pedestrian. In addition, although the estimation process is performed by the method shown in FIG. 3, it is not limited to such a method. , other well-known methods such as template matching.

(2) In the above embodiment, the recognition range is performed after the estimation range is extracted by the estimation process, but the extraction of the estimation range may be omitted. In this case, a captured image may be acquired, and an arbitrary range may be sequentially extracted as an estimated range from the captured image range acquired, and recognition processing may be performed.

(3) In the above embodiment, the distribution pattern of a part of poles installed on the road was given as an example of the object pattern. Patterns representing portions of other objects may also be stored, such as the outline shape of the .

(4) As an example of the object pattern, a part of the object recognized as the leg portion of the pedestrian is given as an example, but any portion that can be easily recognized as a pedestrian is not limited to the leg portion, such as the head. A portion of the object that is erroneously recognized as the object pattern may be included. Also, a distribution pattern corresponding to the contour shape of the head may be set as the distribution pattern to be compared with the attention range. Specifically, for example, a distribution pattern may be set in which the edge degree is high in the center and the edge degree is low in the periphery along the horizontal direction.

(5) In the above embodiment, the edge degree is obtained by calculating the number of edge points for each boundary of pixel columns arranged in the vertical direction in the selection range. The degree of edging is not limited to being calculated, and for example, the degree of edging may be calculated by multiplying the luminance difference of each edge point and then adding them. The edge degree may also be calculated by adding values obtained by applying a Sobel filter to the luminance values of the image for each boundary of pixel rows arranged in the vertical direction.

(6) In the above embodiment, the pedestrian recognition process ends when the estimated range does not exist in the captured image in the pedestrian recognition process, but the process is not limited to such a process. For example, a process of notifying that it is estimated that a pedestrian does not exist in the captured image may be performed.

(7) Furthermore, although the object pattern is stored in the distribution pattern memory 64, the object pattern stored in the distribution pattern memory 64 may be updated. For example, a part of an object recognized as a pedestrian in the estimation process but determined not to be a pedestrian in the recognition process is selected, and the part of the selected object is added to the distribution pattern memory 64 as an object pattern. You may have the structure which carries out.

(8) Further, it may have a sensor or sonar for detecting a target. Alternatively, the range of the target detected by the sensor or sonar may be extracted as the estimated range and the recognition process may be executed.

(9) Further, a sensor for detecting the direction in which the vehicle has moved and the amount of movement of the vehicle may be provided. The amount of movement of the target in the image may be calculated from the direction and amount of movement of the own vehicle detected by the sensor, and the target that has moved along with the movement may be detected.

(10) Furthermore, the camera module may be, for example, a camera module with a wide-angle lens. In this case, the image processing device may have a conversion unit that converts the wide-angle image into a normal image when acquiring the image. Note that the normal image is not an image captured by a wide-angle lens or the like, but an image similar to the one viewed with normal human eyes.

(11) In the above embodiment, the distribution patterns include object patterns representing objects, but pedestrian patterns representing pedestrians may also be stored.
(12) In the above embodiment, the object patterns are sequentially called in the recognition process, and when there is no object pattern that has not been called, the estimated range is recognized as representing a pedestrian. It is not limited to such steps. For example, conversely, pedestrian patterns may be invoked in sequence, and the estimated range may be recognized as representing an object when there are no more uninvoked pedestrian patterns.

(14) Alternatively, both the object pattern and the pedestrian pattern may be called, and among the called patterns, the pattern with the highest degree of matching may be recognized as part of the target represented by the selection range.

(15) In the above embodiment, the recognition processing is used for pedestrian recognition, but it is not limited to such processing. For example, it may be used to recognize targets other than pedestrians. Also, the estimation process may not be performed. When the estimation process is not executed, predetermined ranges in the captured image may be sequentially extracted as the estimated range or the attention range.

(16) In the above embodiment, the pedestrian range and the object range are output in the recognition process. However, the recognition result may not include both the pedestrian range and the object range. For example, the pedestrian range may be output without outputting the object range as the recognition result.

(17) A plurality of functions possessed by one component in the above embodiment may be realized by a plurality of components, or a function possessed by one component may be realized by a plurality of components. . Also, a plurality of functions possessed by a plurality of components may be realized by a single component, or a function realized by a plurality of components may be realized by a single component. Also, part of the configuration of the above embodiment may be omitted. Moreover, at least part of the configuration of the above embodiment may be added or replaced with respect to the configuration of the other above embodiment. It should be noted that all aspects included in the technical idea specified by the wording described in the claims are embodiments of the present disclosure.

(18) In addition to the above-described image processing device, a system having the image processing device as a component, a program for making a computer function as the image processing device, a non-transitional actual record such as a semiconductor memory in which this program is recorded The present disclosure can also be implemented in various forms such as media and image processing methods.

REFERENCE SIGNS LIST 1 image recognition system 10 camera module 20 control device 30 brake actuator 40 display device 50 speaker 60 image processing device 61 CPU 62 processing memory 63 matching pattern memory , 64... Distribution pattern memory, 100... Captured image, 101... First pedestrian, 101a... First estimated range, 101b... First attention range, 102... Second pedestrian, 102a... Second estimated range, 102b... Third Second range of interest, 103... Pole, 103a... Third estimated range, 103b... Third range of interest.

Claims (6)

an image acquisition unit (S210) configured to acquire a captured image (100) captured by an imaging device (10) mounted on a vehicle;
a range setting unit (S110, S310) configured to set, as a target range, a range in which a target is estimated to exist in the captured image acquired by the image acquisition unit;
An edge configured to be detected as an edge point representing a contour portion of the target along the horizontal direction in the captured image in at least part of the target range set by the range setting unit. a detection unit (S340);
An edge distribution in which edge degrees, which are numerical values obtained by adding the number of edge points calculated by the edge detection unit in the vertical direction of the captured image, are arranged along the horizontal direction of the captured image. a distribution calculator (S370) configured to calculate
a distribution pattern memory (64) configured to store at least one distribution pattern, which is a representative pattern of the edge distribution previously associated with at least a part of the target, for each type of the target;
According to the degree of similarity between the edge distribution calculated by the distribution calculating unit and each of at least one of the distribution patterns stored in the distribution pattern memory, the target range determines the target corresponding to the distribution pattern. a target determination unit (S380) configured to determine whether or not the
An image processing device comprising: The image processing device according to claim 1,
The image processing device, wherein the distribution pattern memory stores, as the distribution pattern, an object pattern, which is a pattern representing an object other than a pedestrian, that is a target object. The image processing device according to claim 2,
The range setting unit is configured to set the target range estimated to represent a pedestrian as an estimated range,
A pedestrian determination unit ( S130, S400). The image processing device according to claim 2 or 3,
The image processing device, wherein the object patterns include a pole pattern that is a pattern representing a pole that is a columnar body installed on a road. The image processing device according to any one of claims 1 to 4,
The image processing device that determines the target based on the number of peaks of the edge degree in the edge distribution. The image processing device according to any one of claims 1 to 5,
The image processing apparatus uses, as the edge point, a point at which an absolute value of a luminance difference between pixels adjacent in the horizontal direction is equal to or greater than a predetermined luminance threshold.

Images