JP2020144758A - Moving object detector, moving object detection method, and computer program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the accuracy of detecting a moving object reflected in an image. A feature point detection unit (20) extracts feature points in an image captured by an image pickup device (12) and detects movements of a plurality of feature points over a plurality of images. The object detection unit 24 detects a moving object reflected in an image by grouping a plurality of feature points having similar movements. The tracking unit 26 tracks the movement of a moving object over a plurality of images. When a plurality of candidate regions of a moving object are detected by grouping a plurality of feature points in a certain image, the object detection unit 24 has a feature point among the pixels around the feature point for each of the plurality of feature points. Extract a similar area in which areas of pixels similar to are communicated with each other. The object detection unit 24 synthesizes a plurality of candidate regions so as to include similar regions, and detects the combined region as a region of a moving object reflected in the certain image. [Selection diagram] Fig. 3

Description

The present disclosure relates to data processing techniques, in particular to moving object detection devices, moving object detection methods and computer programs.

As a method of detecting a moving object (also referred to as a "moving object") reflected in an image, a motion vector of a feature point over a plurality of images is obtained, and a plurality of feature points are clustered based on the motion vector of each feature point. By doing so, there is a method of detecting the area of a moving object on the image.

International Publication No. 2017/038123

In the conventional moving object detection method, a plurality of moving objects are actually detected from an image showing one moving object, and as a result, the moving object may not be tracked correctly.

The present disclosure has been made in view of these problems, and one object is to improve the detection accuracy of moving objects reflected in an image.

In order to solve the above problems, the moving object detection device of an embodiment of the present invention extracts feature points in an image captured by the image pickup device, and has a plurality of feature points over a plurality of images captured at different times. A feature point detection unit that detects movement, an object detection unit that detects a moving object that appears in an image by grouping a plurality of feature points with similar movements detected by the feature point detection unit, and a plurality of images. , A tracking unit that tracks the movement of a moving object detected by the object detecting unit. When a plurality of candidate regions for moving objects are detected by grouping a plurality of feature points in a certain image, the object detection unit sets each of the plurality of feature points as a feature point among the pixels around the feature point. A plurality of candidate regions are combined so as to include a similar region extraction unit that extracts a similar region in which similar pixel regions are communicated and a similar region extracted by the similar region extraction unit, and the combined region is combined into a certain image. Includes a compositing unit that detects as an area of a moving object that appears.

Another aspect of the present invention is a moving object detection method. This method extracts feature points in an image captured by an imaging device and detects the movement of a plurality of feature points over a plurality of images captured at different times, and a process of detecting the movement of a plurality of feature points having similar movements. A computer executes a process of detecting a moving object reflected in an image by grouping and a process of tracking the movement of the moving object over a plurality of images, and a plurality of processes of detecting the moving object are performed in a certain image. When a plurality of candidate regions for moving objects are detected by grouping the feature points of, for each of the plurality of feature points, the regions of the pixels around the feature points that are similar to the feature points are communicated with each other. It includes a process of extracting a region and a process of synthesizing a plurality of candidate regions so as to include the extracted similar regions and detecting the combined region as a region of a moving object appearing in a certain image.

It should be noted that any combination of the above components and the expression of the present disclosure converted between a system, a computer program, a recording medium on which a computer program is recorded, and the like are also effective as aspects of the present disclosure.

According to the present disclosure, it is possible to improve the detection accuracy of a moving object appearing in an image.

It is a figure which shows the conventional example of the moving object detection in a vehicle. 2 (a) -FIG. 2 (f) is a diagram showing an example of detecting a moving object. It is a figure which shows the structure of the driving support system of an Example. 4 (a)-(d) are diagrams showing an example of extracting similar regions. 5 (a)-(c) are diagrams showing an example of synthesizing detection frames.

FIG. 1 shows a conventional example of detecting a moving object in a vehicle. First, feature points (corner points, etc.) in the image (in other words, each frame of the image) captured by the imaging device are detected. Next, the feature points are tracked by collating the feature points detected in the current frame with the feature points detected in the previous frame, and the motion vector of the feature points is obtained (Example 1). The motion vector can be said to be a motion vector, and includes a travel distance and a travel direction.

Next, a motion vector of a group of feature points (hereinafter, also referred to as "feature point cloud") in which a plurality of feature points in a neighborhood range and having the same motion vector are grouped (also referred to as clustering). Is calculated (Example 2). Next, a plurality of feature point groups that are in the neighborhood range and have the same motion vector are grouped. A frame (which can be said to be an area) surrounding the grouped results is output as a "detection frame" for a moving object (Example 3).

The movement of the detection frame over multiple frames is tracked in chronological order. Specifically, the movement of the detection frame is tracked using elements that can identify the detection frame, such as the width and height of the detection frame and the coordinates in the image (Example 4). Based on the tracked movement of the detection frame, the movement direction, movement distance, and movement speed of the detection frame in the next frame are predicted, and the collision prediction is based on the prediction result and the own vehicle position and the own vehicle speed. Calculate the time (Time-To-Collision, TTC) (Example 5). By inputting the TTC to the driving control device, vehicle control (emergency braking, etc.) according to the TTC is executed.

FIG. 2A-FIG. 2F shows an example of detecting a moving object. FIGS. 2 (a) and 2 (f) include an image 6 captured by an in-vehicle camera, and show in chronological order how another vehicle (that is, a moving object) is gradually approaching. As shown in FIGS. 2 (c) and 2 (e), if the image of the other vehicle can be detected by one detection frame 7, the movement of the other vehicle can be accurately tracked.

However, as shown in FIGS. 2 (d) and 2 (g), if the detection frame 7 for the image of the other vehicle is divided into a plurality of parts, the accuracy of tracking the movement of the other vehicle is lowered. Further, as shown in FIGS. 2 (f) and 2 (g), when the feature points of another vehicle and the feature points of the background are clustered into one group, the size of the detection frame 7 becomes inaccurate. After all, the accuracy of tracking the movement of other vehicles is reduced.

In order to solve such a problem, in the embodiment, the moving object detection that improves the detection accuracy of the moving object by integrating a plurality of detection frames detected for one moving object shown in the image into one detection frame. Propose a device.

FIG. 3 shows the configuration of the driving support system 10 of the embodiment. The driving support system 10 is a data processing system mounted on a vehicle, and includes an imaging device 12, a moving object detection device 14, and a driving control device 16. These devices are connected via a known in-vehicle network such as CAN (Controller Area Network), Ethernet (registered trademark), and MOST (Media Oriented Systems Transport).

The image pickup device 12 takes an image of the surroundings of the vehicle and outputs an image of the image pickup result. The moving object detection device 14 detects a moving object reflected in the image output from the image pickup device 12. The driving control device 16 controls the behavior of the vehicle based on the information about the moving object output from the moving object detecting device 14. The moving object detection device 14 and the operation control device 16 may be mounted as an ECU (Electronic Control Unit). Although not shown in FIG. 3, the operation control device 16 may be connected to a plurality of ECUs that control the engine, brakes, steering, and the like.

In the embodiment, the information about the moving object output by the moving object detecting device 14 includes the TTC for the moving object. The operation control device 16 activates the automatic emergency brake (Autonomous Emergency Braking, AEB) when the TTC is smaller than a predetermined threshold value, specifically, when the calculated TTC time is shorter than the normal stop time. The above-mentioned predetermined threshold value may be the time required from the start of operation of the brake to the stop position (completion of stop) when the vehicle is stopped without activating the automatic emergency brake while the vehicle is running.

FIG. 3 includes a block diagram showing a functional block of the moving object detection device 14. The moving object detection device 14 includes a feature point detection unit 20, a feature point buffer 22, an object detection unit 24, a tracking unit 26, and a TTC calculation unit 28.

Each block shown in the block diagram of the present disclosure can be realized by an element or a mechanical device such as a CPU and a memory of a computer in terms of hardware, and can be realized by a computer program or the like in terms of software. , Draws a functional block realized by their cooperation. It is understood by those skilled in the art that these functional blocks can be realized in various forms by combining hardware and software. For example, a computer program including a plurality of modules corresponding to a plurality of functional blocks of the moving object detection device 14 may be stored in the ROM of the moving object detection device 14. The CPU of the driving support system 10 may exert the functions of the plurality of functional blocks by reading the computer program into the RAM and executing the program.

The feature point detection unit 20 extracts feature points in an image (which can be said to be one frame of an image) output from the image pickup apparatus 12. The feature point detection unit 20 may extract feature points by using a known edge detection method or corner detection method. Further, the feature point detection unit 20 detects motion vectors of a plurality of feature points over a plurality of images (in other words, a plurality of images captured at different times) sequentially output from the image pickup apparatus 12.

The feature point buffer 22 stores data (coordinates in the image, feature amount, etc.) related to the feature point for each image output from the image pickup device 12.

The object detection unit 24 acquires a plurality of feature points detected by the feature point detection unit 20 from the feature point buffer 22, and groups the plurality of feature points based on the motion vector of each feature point into an image. Detects reflected moving objects. The object detection unit 24 outputs data related to the detected moving object (for example, data related to the integrated frame described later) to the tracking unit 26. The detailed configuration of the object detection unit 24 will be described later.

The tracking unit 26 tracks the movement of the moving object detected by the object detecting unit 24 over a plurality of images sequentially output from the imaging device 12, and calculates the moving direction, moving distance, and moving speed of the moving object. The tracking unit 26 outputs these data to the TTC calculation unit 28.

The TTC calculation unit 28 estimates the moving direction, moving distance, and moving speed of the moving object at a future point in time based on the moving direction, moving distance, and moving speed of the moving object calculated by the tracking unit 26. The TTC calculation unit 28 calculates the TTC for a moving object based on the estimation result, the own vehicle position, and the own vehicle speed. The TTC calculation unit 28 may acquire the own vehicle position and the own vehicle speed from an in-vehicle sensor or the like via an in-vehicle network. Further, the feature point detection unit 20, the tracking unit 26, and the TTC calculation unit 28 may be realized by a known technique.

The configuration of the object detection unit 24 will be described in detail. The object detection unit 24 includes a clustering unit 30, a similar region extraction unit 32, and a synthesis unit 34. The clustering unit 30 has the same or similar motion vector values (in other words, a predetermined difference) detected by the feature point detection unit 20 in the image output from the image pickup device 12 (also referred to as “target image” here). A candidate region for a moving object is detected by grouping a plurality of feature points (below the threshold value). An appropriate value may be determined for the above threshold value by the knowledge of the developer or an experiment using the driving support system 10.

When a plurality of candidate regions of moving objects are detected from the target image by the clustering unit 30, the similar region extraction unit 32 describes each of the plurality of feature points (also referred to as “target feature points”) around the target feature points. A region (also referred to as a "similar region") in which a region of pixels similar to the target feature point (also referred to as a "similar point") is communicated is extracted from the target image.

Specifically, when searching for a new similarity from the target image, the similarity region extraction unit 32 groups a plurality of pixels in the same direction among the pixels around the target feature point or the detected similarity point. , Calculate the degree of difference between a group of multiple pixels and the target feature point. In the calculation of the degree of difference (in other words, the calculation of the degree of similarity), texture, texture, color, contrast, entropy and the like may be compared. Further, the similar region extraction unit 32 may perform texture analysis of the target image by using a known spatial density level dependence method (SGLDM), density level difference method (GLDM), or density histogram method (GLHM).

FIGS. 4 (a)-(d) show an example of extracting a similar region. Each block (point of each grid) in FIGS. 4 (a)-(d) shows a pixel in the target image. As shown in FIG. 4A, the similar region extraction unit 32 calculates the degree of difference between the four pixel groups located on the top, bottom, left, and right of the target feature point 62 and the target feature point 62. In the embodiment, one pixel group is 3 pixels. The feature amount of the pixel group may be an average value of the feature amount of each pixel. The similarity region extraction unit 32 detects the central pixel of the pixel group whose degree of difference from the target feature point 62 is equal to or less than a predetermined threshold value as a similarity point, and detects the similarity point 64a in FIG. 4A.

FIG. 4B shows the processing following FIG. 4A. The similar region extraction unit 32 calculates the degree of difference between the four pixel groups located on the top, bottom, left, and right of the similarity point 64a and the target feature point 62. In FIG. 4B, the similarity region extraction unit 32 identifies a pixel group located to the right of the similarity point 64a as a pixel group whose degree of difference from the target feature point 62 is equal to or less than a predetermined threshold value, and determines the similarity point 64b. To detect.

FIG. 4C shows the processing following FIG. 4B. The similar region extraction unit 32 calculates the degree of difference between the four pixel groups located on the top, bottom, left, and right of the similarity point 64b and the target feature point 62. In FIG. 4C, the similarity region extraction unit 32 identifies pixel groups located above, below, and to the right of the similarity point 64b as pixel groups whose degree of difference from the target feature point 62 is equal to or less than a predetermined threshold value, and the similarity points. 64c, similarity 64d and similarity 64e are detected.

FIG. 4D shows the processing following FIG. 4C. The similar region extraction unit 32 calculates the degree of difference between the four pixel groups located on the top, bottom, left, and right of the similarity point 64e and the target feature point 62. In FIG. 4D, the similarity region extraction unit 32 identifies a pixel group located to the right of the similarity point 64e as a pixel group whose degree of difference from the target feature point 62 is equal to or less than a predetermined threshold value, and determines the similarity point 64f. To detect.

In this way, the similar region extraction unit 32 repeats the search for similar points with respect to the target feature point 62. If any of the four pixel groups located on the top, bottom, left, and right of a certain similarity point (for example, the similarity point 64f) has a degree of difference from the target feature point 62 larger than a predetermined threshold value, the similarity area extraction unit 32 has the similarity (for example, the similarity point For example, the search starting from the similarity point 64f) is completed. The threshold value of the degree of difference used in the explanation related to FIG. 4 may be appropriately determined by the knowledge of the developer and the experiment using the driving support system 10.

The similar region extraction unit 32 extracts a region in which one or more similar points detected by the search are communicated with each other, starting from the target feature point 62, as a similar region. In the example of FIGS. 4A-(d), the region connecting the target feature points 62, the similarity 64a, the similarity 64b, the similarity 64c, the similarity 64d, the similarity 64e, and the similarity 64f is extracted as the similarity region. To do.

Returning to FIG. 3, when the similar region extracted by the similar region extraction unit 32 extends over the plurality of candidate regions detected by the clustering unit 30, the synthesis unit 34 sets a plurality of candidate regions so as to include the similar region. Synthesize. The synthesizing unit 34 detects the synthesized area as the area of the moving object reflected in the target image, and outputs information about the area of the moving object to the tracking unit 26. The information about the area of the moving object may be the information of the coordinates occupied by the area of the moving object in the target image.

FIGS. 5 (a)-(c) show an example of synthesizing detection frames. FIG. 5A shows a detection frame detected by the clustering unit 30 from the target image 40. The figure shows that the detection frame 42 and the detection frame 44 were detected from one vehicle. FIG. 5B shows a similar region 46 and a similar region 48 extracted from the target image 40 by the similar region extraction unit 32.

Both the similar region 46 and the similar region 48 straddle both the detection frame 42 and the detection frame 44. Therefore, as shown in FIG. 5C, the synthesis unit 34 synthesizes the detection frame 42 and the detection frame 44 so as to include both the similar region 46 and the similar region 48, and combines the integrated frame 50, which is the result of the synthesis. Generate. The synthesis unit 34 outputs information (coordinate information, etc.) indicating the position of the integrated frame 50 in the target image 40 to the tracking unit 26 as information indicating the region of the moving object reflected in the target image 40.

When detecting the region of the moving object reflected in the target image 40, the synthesis unit 34 excludes the similar region from the region of the moving object when the similar region extracted by the similar region extraction unit 32 is smaller than a predetermined size. In the example of FIG. 5C, a similar region having a predetermined size or larger is not extracted from the right side portion (that is, the portion where the background is reflected) of the detection frame 44. Therefore, the synthesis unit 34 excludes the right side portion of the detection frame 44 from the integration frame 50.

An appropriate value may be determined for the predetermined size (that is, the threshold value) based on the knowledge of the developer and an experiment using the driving support system 10. Further, the synthesis unit 34 may delete the information of the feature points not included in the integration frame 50, or suppress the output of the information of the feature points not included in the integration frame 50 to the tracking unit 26. May be good.

The operation of the driving support system 10 with the above configuration will be described with reference to FIG.
The image pickup device 12 mounted on the vehicle outputs an image showing the surroundings of the vehicle to the moving object detection device 14. The feature point detection unit 20 detects feature points for each frame of the image output from the image pickup apparatus 12, and detects a motion vector of each feature point. The feature point buffer 22 stores information on feature points for each frame of the video.

The clustering unit 30 detects a candidate region of a moving object reflected in each frame by grouping a plurality of feature points having similar motion vector values for each frame of the video. If there is only one candidate region for the moving object detected by the clustering unit 30, the object detection unit 24 outputs the candidate region as a region for the moving object to the tracking unit 26.

On the other hand, when a plurality of candidate regions for moving objects are detected from a certain frame, the similar region extraction unit 32 extracts similar regions for each feature point from the frame. When the similar region spans a plurality of candidate regions, the synthesizing unit 34 synthesizes the plurality of candidate regions so as to include the similar region, and detects the synthesized region as a region of a moving object reflected in the frame. Further, when the extracted similar region is smaller than a predetermined size, the synthesis unit 34 excludes the similar region from the region of the moving object. The object detection unit 24 outputs the region of the moving object detected by the synthesis unit 34 to the tracking unit 26.

The tracking unit 26 tracks the movement of a moving object over a plurality of frames. The TTC calculation unit 28 estimates the movement of the moving object at a future time point (for example, the next frame) based on the movement of the moving object over a plurality of frames. The TTC calculation unit 28 calculates the TTC for the moving object based on the estimation result and outputs it to the operation control device 16. The driving control device 16 controls the behavior of the vehicle based on the TTC, and activates, for example, an emergency brake for collision avoidance.

According to the moving object detection device 14 of the embodiment, the accuracy of detecting the moving object reflected in the image can be improved, in other words, the region of the moving object can be correctly detected from the image. For example, it is possible to prevent one moving object from being detected as a plurality of moving objects. As a result, the movement of the moving object in the time series can be correctly detected, and the accuracy of estimating the movement of the moving object at a future point in time can be improved.

Further, according to the moving object detection device 14, by excluding a small similar region from the region of the moving object, it is possible to prevent the feature point region (background region or the like) other than the moving object from being mistakenly detected as a moving object. As a result, the detection accuracy of moving objects can be further improved. Furthermore, in the extraction of similar regions, a group of pixels instead of one pixel is compared with the feature points. As a result, the amount of calculation for similarity determination (in other words, difference degree determination) can be reduced.

The present disclosure has been described above based on the examples. This embodiment is an example, and it will be understood by those skilled in the art that various modifications are possible for each component of the embodiment or a combination of each processing process, and that such modifications are also within the scope of the present disclosure. is there.

The moving object detection device 14 of the above embodiment tracks the movement of the moving object in time series, calculates the TTC for the moving object based on the tracking result, and outputs the TTC to the operation control device 16. As a modification, the moving object detection device 14 may output the result of tracking the movement of the moving object in time series to an external device, and the external device may output predetermined data based on the tracking result such as calculation of TTC. The process may be executed.

The moving object detection device 14 of the above embodiment was mounted on a vehicle. As a modification, the moving object detection device 14 (or the function of the moving object detection device 14) may be mounted on a device (server or the like) that is a device other than the vehicle and executes data processing related to the moving object. Further, the moving object detection device 14 accepts an image captured by an image pickup device 12 (for example, a surveillance camera fixed to a building) installed outside the device equipped with the moving object detection device 14 via a communication network. Alternatively, a moving object may be detected from the image.

Any combination of the examples and modifications described above is also useful as an embodiment of the present disclosure. The new embodiments resulting from the combination have the effects of the combined examples and the modifications. It is also understood by those skilled in the art that the functions to be fulfilled by each of the constituent elements described in the claims are realized by a single component or a cooperation thereof shown in the examples and modifications.

The techniques described in the examples and modifications may be specified by the following items.
[Item 1]
A feature point detection unit that extracts feature points in an image captured by an imaging device and detects the movement of a plurality of feature points over a plurality of images captured at different times.
An object detection unit that detects a moving object reflected in the image by grouping a plurality of feature points having similar movements detected by the feature point detection unit.
A tracking unit that tracks the movement of a moving object detected by the object detection unit over the plurality of images, and a tracking unit.
With
The object detection unit
When a plurality of candidate regions for moving objects are detected by grouping a plurality of feature points in a certain image, for each of the plurality of feature points, among the pixels around the feature points, pixels similar to the feature points A similar area extraction unit that extracts similar areas that communicate the areas,
It includes a compositing unit that synthesizes a plurality of candidate regions so as to include the similar regions extracted by the similar region extraction unit and detects the synthesized region as a region of a moving object appearing in the certain image.
Moving object detector.
[Item 2]
When the similar region extracted by the similar region extraction unit is smaller than a predetermined size, the synthesis unit excludes the similar region from the region of the moving object.
The moving object detection device according to item 1.
[Item 3]
The similar region extraction unit groups a plurality of pixels in the same direction among the pixel around the feature point or the similar pixel, and calculates the degree of difference between the group of the plurality of pixels and the feature point.
The moving object detection device according to item 1 or 2.
[Item 4]
The image pickup device is used in a vehicle.
The moving object detection device according to item 1.
[Item 5]
A process of extracting feature points in an image captured by an imaging device and detecting the movement of a plurality of feature points over a plurality of images captured at different times.
A process of detecting a moving object appearing in the image by grouping a plurality of feature points having similar movements, and
The process of tracking the movement of a moving object over a plurality of images, and
The computer runs,
The process of detecting the moving object is
When a plurality of candidate regions for moving objects are detected by grouping a plurality of feature points in a certain image, for each of the plurality of feature points, among the pixels around the feature points, pixels similar to the feature points The process of extracting similar areas that communicate the areas, and
A process of synthesizing a plurality of candidate regions so as to include an extracted similar region and detecting the combined region as a region of a moving object reflected in the certain image is included.
Moving object detection method.
[Item 6]
A process of extracting feature points in an image captured by an imaging device and detecting the movement of a plurality of feature points over a plurality of images captured at different times.
A process of detecting a moving object appearing in the image by grouping a plurality of feature points having similar movements, and
The process of tracking the movement of a moving object over a plurality of images, and
Let the computer run
The process of detecting the moving object is
When a plurality of candidate regions for moving objects are detected by grouping a plurality of feature points in a certain image, for each of the plurality of feature points, among the pixels around the feature points, pixels similar to the feature points The process of extracting similar areas that communicate the areas, and
A process of synthesizing a plurality of candidate regions so as to include an extracted similar region and detecting the combined region as a region of a moving object reflected in the certain image is included.
Computer program.

10 Driving support system, 12 Imaging device, 14 Moving object detection device, 20 Feature point detection section, 24 Object detection section, 26 Tracking section, 32 Similar area extraction section, 34 Synthesis section.

Claims (6)

A feature point detection unit that extracts feature points in an image captured by an imaging device and detects the movement of a plurality of feature points over a plurality of images captured at different times.
An object detection unit that detects a moving object reflected in the image by grouping a plurality of feature points having similar movements detected by the feature point detection unit.
A tracking unit that tracks the movement of a moving object detected by the object detection unit over the plurality of images, and a tracking unit.
With
The object detection unit
When a plurality of candidate regions for moving objects are detected by grouping a plurality of feature points in a certain image, for each of the plurality of feature points, among the pixels around the feature points, pixels similar to the feature points A similar area extraction unit that extracts similar areas that communicate the areas,
It includes a compositing unit that synthesizes a plurality of candidate regions so as to include the similar regions extracted by the similar region extraction unit and detects the synthesized region as a region of a moving object appearing in the certain image.
Moving object detector. When the similar region extracted by the similar region extraction unit is smaller than a predetermined size, the synthesis unit excludes the similar region from the region of the moving object.
The moving object detection device according to claim 1. The similar region extraction unit groups a plurality of pixels in the same direction among the pixel around the feature point or the similar pixel, and calculates the degree of difference between the group of the plurality of pixels and the feature point.
The moving object detection device according to claim 1 or 2. The image pickup device is used in a vehicle.
The moving object detection device according to claim 1. A process of extracting feature points in an image captured by an imaging device and detecting the movement of a plurality of feature points over a plurality of images captured at different times.
A process of detecting a moving object appearing in the image by grouping a plurality of feature points having similar movements, and
The process of tracking the movement of a moving object over a plurality of images, and
The computer runs,
The process of detecting the moving object is
When a plurality of candidate regions for moving objects are detected by grouping a plurality of feature points in a certain image, for each of the plurality of feature points, among the pixels around the feature points, pixels similar to the feature points The process of extracting similar areas that communicate the areas, and
A process of synthesizing a plurality of candidate regions so as to include an extracted similar region and detecting the combined region as a region of a moving object reflected in the certain image is included.
Moving object detection method. A process of extracting feature points in an image captured by an imaging device and detecting the movement of a plurality of feature points over a plurality of images captured at different times.
A process of detecting a moving object appearing in the image by grouping a plurality of feature points having similar movements, and
The process of tracking the movement of a moving object over a plurality of images, and
Let the computer run
The process of detecting the moving object is
When a plurality of candidate regions for moving objects are detected by grouping a plurality of feature points in a certain image, for each of the plurality of feature points, among the pixels around the feature points, pixels similar to the feature points The process of extracting similar areas that communicate the areas, and
A process of synthesizing a plurality of candidate regions so as to include an extracted similar region and detecting the combined region as a region of a moving object reflected in the certain image is included.
Computer program.

Images