Object detection system, object detection method, and object detection program

Abstract

To provide an object detection system capable of detecting a target object from an image at high speed.SOLUTION: Object presence region prediction means 81 predicts an object presence region in which a target object is present in a current image from information indicative of a target object detected from a past image. Object presence region fragment generation means 82 generates an object presence region fragment, which is a partial region of the object presence region, from the object presence region. Object detection means 83 detects an object detection fragment, which is a region including the target object, from the object presence region fragment. Target object detection means 84 detects a target object from the current image using the object detection fragment.SELECTED DRAWING: Figure 10

Description

Technical Field

[0001] The present invention relates to an object detection system, an object detection method, and an object detection program for detecting a target object in an image.

Background Art

[0002] An example of a method for detecting an object is described in Non-Patent Document 1. In the method described in Non-Patent Document 1, based on the position of a target object detected in a certain past image, a region where the target object is assumed to exist in the current image is set, and the position and size of the object within that region are determined by object detection.

[0003] Specifically, in the method described in Non-Patent Document 1, the approximate position and size of the region containing the target object are predicted by setting a region where the target object is assumed to exist. Then, the position and size of the target object are determined for a part of the image obtained by prediction, and the result is used as the object detection result.

Prior Art Documents

Non-Patent Documents

[0004]

Non-Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] Generally, object detection is computationally intensive, so to speed up processing, it is desirable to make the image to be processed as small as possible. Therefore, in order to reduce the computational load, a method can be considered in which processing is performed on only a portion of the image in which the target object is to be detected, as described in Non-Patent Document 1.

[0006] On the other hand, for example, if the object to be detected is large, the method described in Non-Patent Document 1 will end up repeatedly performing object detection on the large object, and it is difficult to say that the computational load can be sufficiently reduced. Therefore, it is difficult to quickly estimate the target object from an image.

[0007] Therefore, the present invention aims to provide an object detection system, an object detection method, and an object detection program that can quickly detect target objects from images. [Means for solving the problem]

[0008] The object detection system according to the present invention is characterized by comprising: an object existence region prediction means that predicts an object existence region, which is the region in which a target object exists in the current image, from information indicating a target object detected from past images; an object existence region fragment generation means that generates an object existence region fragment, which is a sub-region of the object existence region, from the object existence region; an object detection means that detects an object detection fragment, which is the region containing the target object, from the object existence region fragment; and a target object detection means that detects the target object from the current image using the object detection fragment.

[0009] The object detection method according to the present invention is characterized in that a computer predicts an object existence region, which is the region in which a target object exists in the current image, from information indicating a target object detected from past images; the computer generates an object existence region fragment, which is a subregion of the object existence region, from the object existence region fragment; the computer detects an object detection fragment, which is the region containing the target object, from the object existence region fragment; and the computer detects the target object from the current image using the object detection fragment.

[0010] The object detection program according to the present invention is characterized by causing a computer to perform an object existence region prediction process that predicts an object existence region, which is the region in which a target object exists in the current image, based on information indicating a target object detected from past images; an object existence region fragment generation process that generates an object existence region fragment, which is a sub-region of the object existence region, from the object existence region fragment; an object detection process that detects an object detection fragment, which is the region containing the target object, from the object existence region fragment; and a target object detection process that detects a target object from the current image using the object detection fragment. [Effects of the Invention]

[0011] According to the present invention, target objects can be detected from images at high speed. [Brief explanation of the drawing]

[0012] [Figure 1] This is a block diagram showing an example configuration of the first embodiment of the object detection system of the present invention. [Figure 2] This is an explanatory diagram showing examples of object detection results and object presence regions for past images. [Figure 3] This is an explanatory diagram illustrating an example of a process for generating a group of object existence region fragments. [Figure 4] This is an explanatory diagram illustrating an example of the process for detecting object detection fragments. [Figure 5] This is an explanatory diagram illustrating an example of a process for detecting a target object. [Figure 6] This flowchart shows an example of the operation of the object detection system according to the first embodiment. [Figure 7] This block diagram shows an example configuration of a second embodiment of the object detection system of the present invention. [Figure 8] This is an explanatory diagram illustrating an example of a process for predicting the region where an object exists. [Figure 9] This flowchart shows an example of the operation of the object detection system according to the second embodiment. [Figure 10] This is a block diagram outlining the object detection system according to the present invention.

Best Mode for Carrying Out the Invention

[0013] Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the following description, an image for detecting a target object is referred to as a current image. The current image is, for example, an image sequentially captured by a fixed-point camera such as a surveillance camera. Further, in the following description, as a specific example, the case where the target object is a vehicle is illustrated, but the target object is not limited to a vehicle.

[0014] Also, in the present embodiment, it is assumed that the detection of the target object has already been performed from an image captured in the past (hereinafter referred to as a past image) rather than the current image, and information indicating the target object detected from the past image has been calculated. Examples of the information indicating the target object include information indicating the existence region of the target object and an image (hereinafter referred to as a past partial image) in which a portion including the target object is extracted.

[0015] The existence region of the target object is a region including the target object, and is, for example, a rectangular region represented by the top-left vertex coordinates and the width and height. Alternatively, the existence region of the target object may be a rectangular region represented by the top-left vertex coordinates and the bottom-right vertex coordinates.

[0016] [First Embodiment] [Description of Configuration] FIG. 1 is a block diagram showing a configuration example of a first embodiment of an object detection system of the present invention. As shown in FIG. 1, the object detection system 100 of the present embodiment includes an object existence region predictor 200, an object existence region fragment generation unit 300, an object detector 400, a target object detector 500, an imaging device 610, and a storage unit 620.

[0017] The memory unit 620 stores various information necessary for the processing performed by the object detection system 100 of the present embodiment. Also, the memory unit 620 of the present embodiment stores the past image 700 and the past image object detection result 800 described above. The past image object detection result 800 is information indicating the target object detected from the past image. Specifically, it is information indicating the existence region of the target object, or an image in which the portion containing the target object is extracted.

[0018] The object detection system 100 of the present embodiment calculates and outputs an object detection result from the current image and the past image object detection result 800 for the past image 700. In the first embodiment, the case where the information indicating the target object detected from the past image is the information indicating the existence region of the target object will be described.

[0019] The imaging device 610 is a device installed at a predetermined location and captures an image of a detection target. Specifically, the imaging device 610 acquires a current image as a result of imaging. In the present embodiment, it is assumed that the angle of view when the imaging device 610 captures an image does not change over time, and the imaging angles of the current image and the past image are also the same.

[0020] The object existence region predictor 200 predicts the region where the target object exists in the current image (hereinafter referred to as the object existence region) from the information indicating the target object detected from the past image 700 (that is, the past image object detection result 800). The method by which the object existence region predictor 200 predicts the object existence region is arbitrary. The object existence region predictor 200 may, for example, predict the object existence region from the past image object detection result 800 based on a dynamic model such as a Kalman filter.

[0021] Figure 2 is an explanatory diagram showing an example of object detection results and object presence regions for past images. As illustrated in Figure 2, suppose an object detection frame indicating the presence region of the target object is detected from past image 700 as information indicating the target object. In this case, the object presence region predictor 200 uses this object detection frame as information indicating the target object to predict the object presence region of the current image 600. As illustrated in Figure 2, the object presence region may be a rectangular region represented by the coordinates of the top-left vertex, as well as its width and height. Based on the prediction results by the object presence region predictor 200, the object presence region can be said to be an area with a high probability of containing the target object.

[0022] The object existence region fragment generation unit 300 divides the object existence region and generates a sub-region of the object existence region (hereinafter referred to as an object existence region fragment). In doing so, the object existence region fragment generation unit 300 divides the object existence region so that a part of the target object to be detected is included in the object existence region fragment. That is, the object existence region fragment is an image obtained by further dividing a sub-image of the current image 600 obtained from the information of the object existence region, and is an image with a smaller spatial size than the object existence region.

[0023] Furthermore, since the object detector 400, described later, performs object detection processing on the object presence region fragments, the divided target objects should be large enough to be detected by the object detector 400. For this reason, it is preferable that the object presence region fragment generation unit 300 generates object presence region fragments by dividing the object presence region into two equal parts in the vertical or horizontal direction.

[0024] Furthermore, the object existence region fragment generation unit 300 may generate an object existence region fragment to which the position of the object existence region fragment within the object existence region is added. Examples of the position of the object existence region fragment include its position relative to the object existence region before division, for example, information indicating that it was located on the right side of the divided image, or information indicating that it was located at the top. Another example of the position of the object existence region fragment is its relative position based on the coordinates of the top left. By adding such position information, it becomes possible to perform the processing described later (specifically, the processing of detecting the object existence region) with higher accuracy. The processing using this position information will be described later.

[0025] Figure 3 is an explanatory diagram illustrating an example of the process for generating object presence region fragments. When multiple object presence regions are predicted from the current image, the object presence region fragment generation unit 300 generates object presence region fragments from each of the object presence regions. In the example shown in Figure 3, the object presence region fragment generation unit 300 generates object presence region fragment 1200 from object presence region 1100 and object presence region fragment 1210 from object presence region 1110.

[0026] The object detector 400 detects the region containing the target object (hereinafter referred to as the object detection fragment) from the object presence region fragment. The method of representing the object detection fragment is arbitrary. For example, the object detection fragment may be a rectangular region represented by the coordinates of the top-left vertex, as well as its width and height, similar to the object detection result.

[0027] Furthermore, the method by which the object detector 400 of this embodiment detects the region containing the target object (i.e., the object detection fragment) is also arbitrary. In other words, the object detector 400 does not necessarily have to be a special object detector for detecting object detection fragments. The object detector 400 can be any detector that can detect the target object from an image containing a part of the target object. The object detector 400 may be a commonly used object detector such as Yolo (You Look Only Once).

[0028] Figure 4 is an explanatory diagram illustrating an example of the process for detecting object detection fragments. In the example shown in Figure 4, object detection fragments that identify the area of ​​the vehicle were detected from object presence region fragments 1200 and 1210, which both contain a part of the vehicle as the target object.

[0029] The target object detector 500 detects the target object from the current image using the object detection fragment. Specifically, the target object detector 500 calculates the object detection result in the current image 600 from the object detection fragment and the past image object detection result 800 in the past image 700.

[0030] The following describes in detail how the target object detector 500 of this embodiment detects a target object. Figure 5 is an explanatory diagram showing an example of the process of detecting a target object from the current image. In the example shown in Figure 5, it is assumed that the object presence area predictor 200 predicts the object presence area by using an object detection frame indicating the presence area of ​​the target object detected from past images as information indicating the target object. Specifically, it is assumed that the object presence area predictor 200 predicted the object detection frame 1400 and the object detection frame 1300, respectively.

[0031] In this case, the target object detector 500 estimates the object detection frame indicating the area where the target object exists in the current image from the object detection frame and object detection fragments detected from past images. Specifically, the target object detector 500 estimates the horizontal or vertical size of the object detection frame in the current image from the vertical and horizontal sizes (hereinafter referred to as vertical and horizontal sizes) of the object detection frame obtained from past images and the vertical and horizontal sizes of the detection frame obtained from the object detection fragments. The unit of size can be predetermined, such as pixels.

[0032] For example, in Figure 5, suppose the vertical and horizontal dimensions of the object detection frame 1400 obtained from past images were 120 and 100, respectively, and the vertical and horizontal dimensions of the object detection fragment 1700 were 60 and 100, respectively. In this case, since the vertical dimension of the object detection fragment 1700 is 100, the target object detector 500 estimates the horizontal dimension of the object detection frame 1800 in the current image to be 100 * 120 / 100 = 120. This is a rearrangement of the equation 120 / 100 = horizontal dimension of object detection frame 1800 / 100. The objects included in this object detection frame 1800 correspond to the final object detection result.

[0033] Similarly, in Figure 5, suppose the vertical and horizontal dimensions of the object detection frame 1300 obtained from past images were 100 and 110, respectively, and the vertical and horizontal dimensions of the object detection fragment 1600 were 50 and 110, respectively. In this case, since the vertical dimension of the object detection fragment 1500 is 110, the target object detector 500 estimates the horizontal dimension of the object detection frame 1600 in the current image to be 110 * 100 / 110 = 100.

[0034] Here, let's assume that the object presence region fragment generation unit 300 generates object presence region fragments with the position of the object presence region fragment within the object presence region added, as described above. In that case, the target object detector 500 can estimate which part of the object presence region each object detection fragment was located in, and therefore can determine whether to estimate the size of the object detection frame in the vertical or horizontal direction.

[0035] For example, suppose that the object presence region fragment 1210, illustrated in Figure 4, is accompanied by information indicating that it is located in the right half of the divided image. In this case, the object detection fragment 1700, illustrated in Figure 5, can also retain information indicating that it is located in the right half of the divided image. As a result, the target object detector 500 determines that it only needs to calculate the horizontal size of the object detection frame 1800 and then calculate the coordinates of the top-left vertex of the object detection frame.

[0036] In the example shown in Figure 5, the target object detector 500 estimates the coordinates of the top-left vertex of the object detection frame as (x,y) = (x', y coordinate of the top-left vertex of the object detection frame 1800). Here, x' = x coordinate of the top-left vertex of the object detection fragment 1700 - (horizontal size of the object detection frame 1800 - horizontal size of the object detection fragment 1700).

[0037] The target object detector 500 then outputs the detection result of the target object.

[0038] The object presence region predictor 200, the object presence region fragment generation unit 300, the object detector 400, and the target object detector 500 are all implemented by a computer processor (e.g., a CPU (Central Processing Unit) or GPU (Graphics Processing Unit)) that operates according to a program (object detection program).

[0039] For example, the program may be stored in the memory unit 620 of the object detection system 100, and the processor may read the program and operate as the object presence area predictor 200, object presence area fragment generation unit 300, object detector 400, and target object detector 500 according to the program. Alternatively, the functions of the object detection system 100 may be provided in SaaS (Software as a Service) format.

[0040] Furthermore, the object presence region predictor 200, the object presence region fragment generation unit 300, the object detector 400, and the target object detector 500 may each be implemented with dedicated hardware. Also, some or all of the components of each device may be implemented by general-purpose or dedicated circuits, processors, etc., or combinations thereof. These may be configured by a single chip or by multiple chips connected via a bus. Some or all of the components of each device may be implemented by a combination of the above-mentioned circuits, etc., and programs.

[0041] Furthermore, if some or all of the components of the object detection system 100 are implemented by multiple information processing devices or circuits, these multiple information processing devices or circuits may be centrally located or distributed. For example, the information processing devices or circuits may be implemented in a form in which each is connected via a communication network, such as a client-server system or a cloud computing system.

[0042] [Explanation of operation] Next, an example of the operation of the object detection system of this embodiment will be described. Figure 6 is a flowchart showing an example of the operation of the object detection system 100 of this embodiment.

[0043] The object presence region predictor 200 receives the current image and the object detection results for past images (step S1). That is, the object presence region predictor 200 receives information indicating the target object detected from past images as an object detection result. The object presence region predictor 200 predicts the object presence region for the current image based on the object detection results for past images (step S2). The object presence region fragment generation unit 300 generates object presence region fragments from the object presence region (step S3). The object detector 400 performs object detection on the object presence region fragment group and calculates the object detection fragment group (step S4). That is, the object detector 400 detects object detection fragments from the object presence region fragments. Then, the target object detector 500 estimates the object detection result from the object detection fragment group and the object detection results for past images and uses it as the object detection result for the current image (step S5). That is, the target object detector 500 detects the target object from the current image using the object detection fragments.

[0044] [Explanation of effects] Next, the effects of this embodiment will be explained. As described above, in this embodiment, the object presence region predictor 200 predicts the object presence region from information indicating the target object detected from past images, and the object presence region fragment generation unit 300 generates object presence region fragments from the object presence region. Then, the object detector 400 detects object detection fragments from the object presence region fragments, and the target object detector 500 uses the object detection fragments to detect the target object from the current image. Therefore, the target object can be detected from the image at high speed.

[0045] In other words, the object detection system 100 of this embodiment performs object detection using only one of the fragments of the object existence region obtained by dividing the object existence region (i.e., without using the other fragment), rather than the entire object existence region. This enables high-speed inference and shortens the inference time for object detection, i.e., allows for high-speed computation. This is because the spatial size of the image used when detecting the target object is reduced. Furthermore, since the object detection system 100 of this embodiment estimates the object detection result using the object detection fragment, it can output an object detection result that includes the complete target object.

[0046] [Second Embodiment] [Explanation of the structure] Next, a second embodiment of the object detection system according to the present invention will be described. In the second embodiment, the case in which the information indicating the target object detected from past images is an image from which the portion containing the target object has been extracted (i.e., a past partial image) will be described.

[0047] As shown in Figure 7, the object detection system 110 of this embodiment includes an object presence region predictor 210, an object presence region fragment generation unit 300, an object detector 400, a target object detector 500, a camera 610, a storage unit 620, and a past partial image generator 1000. In other words, the object detection system 110 of this embodiment differs from the object detection system 100 of the first embodiment in that it further includes a past partial image generator 1000 and replaces the object presence region predictor 200 with an object presence region predictor 210. The other configurations are the same as those of the first embodiment.

[0048] The past partial image generator 1000 generates past partial images from past images and object detection results for past images. As described above, a past partial image is an image in which the portion containing the target object is extracted from the past image. The method by which the past partial image generator 1000 generates past partial images is arbitrary, and any known object detection method may be used.

[0049] The object location predictor 210 predicts the object location using past partial images as information indicating the target object. Specifically, the object location predictor 210 predicts the object location based on the correlation between past partial images and the current image.

[0050] Figure 8 is an explanatory diagram illustrating an example of the process for predicting the object presence region. In the example shown in Figure 8, the object presence region predictor 210 calculates the correlation between the past partial image 710 and the object contained in the current image 600, using the pixel values ​​of the image as a classical method for calculating the correlation coefficient. Specifically, in the example shown in Figure 8, the object presence region predictor 210 predicts the object presence region by calculating multiple correlations between the past partial image and the part of the current image corresponding to its position, while sliding the past partial image relative to the current image. The object presence region predictor 210 may, for example, predict the current image at a position where a correlation exceeding a predetermined value has been calculated as the object presence region.

[0051] Alternatively, for example, the object location predictor 210 may calculate the correlation between all candidate object location regions in the current image and past partial images, and predict the candidate with the highest correlation as an object location region.

[0052] Alternatively, the object location predictor 210 may use a deep learning model that takes two images as input and outputs the location with the highest correlation between the two images. For example, the Siam (Siamese) network is one such deep learning model. In this case, the object location predictor 210 may input past partial images and the current image into the deep learning model and predict the object location based on the output result.

[0053] The past partial image generator 1000, the object presence region predictor 210, the object presence region fragment generation unit 300, the object detector 400, and the target object detector 500 are all implemented by a computer processor (e.g., a CPU (Central Processing Unit) or GPU (Graphics Processing Unit)) that operates according to a program (object detection program).

[0054] [Explanation of operation] Next, an example of the operation of the object detection system of this embodiment will be described. Figure 9 is a flowchart showing an example of the operation of the object detection system 110 of this embodiment.

[0055] The past partial image generator 1000 receives past images and object detection results for past images (step S11) and generates past partial images (step S12). The object presence region predictor 210 predicts the object presence region using the past partial images (step S13). Subsequent processing is the same as the processing from step S3 onwards illustrated in Figure 6.

[0056] [Explanation of effects] Next, the effects of this embodiment will be described. As described above, in this embodiment, the object presence region predictor 210 predicts the object presence region based on the correlation between past partial images and the current image. Therefore, similar to the first embodiment, the target object can be detected from the image at high speed.

[0057] Next, an overview of the present invention will be described. Figure 10 is a block diagram illustrating an overview of the object detection system according to the present invention. The object detection system 80 according to the present invention includes: an object existence region prediction means 81 (e.g., an object existence region predictor 200) that predicts an object existence region, which is the region in which a target object exists in the current image, from information indicating a target object detected from past images; an object existence region fragment generation means 82 (e.g., an object existence region fragment generation unit 300) that generates an object existence region fragment, which is a subregion of the object existence region, from the object existence region; an object detection means 83 (e.g., an object detector 400) that detects an object detection fragment, which is the region containing the target object, from the object existence region fragment; and a target object detection means 84 (e.g., a target object detector 500) that detects a target object from the current image using the object detection fragment.

[0058] Such a configuration allows for high-speed detection of target objects from images.

[0059] Furthermore, the object presence region prediction means 81 may predict the object presence region by using an object detection frame indicating the presence region of the target object detected from past images as information indicating the target object, and the target object detection means 84 may estimate an object detection frame indicating the presence region of the target object in the current image from the object detection frame and object detection fragments.

[0060] In this case, the target object detection means 84 may estimate the horizontal or vertical size of the object detection frame in the current image from the vertical and horizontal sizes of the object detection frame obtained from past images and the vertical and horizontal sizes of the detection frame obtained from object detection fragments.

[0061] Furthermore, the object existence region fragment generation means 82 may generate an object existence region fragment to which the position of the object existence region fragment within the object existence region has been added.

[0062] Specifically, the object existence region fragment generation means 82 may generate an object existence region fragment in which the position of the object existence region fragment is added to the position of the object existence region fragment relative to the object existence region before division.

[0063] Furthermore, the object existence region fragment generation means 82 may generate object existence region fragments by dividing the object existence region into two equal parts in the vertical or horizontal direction.

[0064] Alternatively, the object presence region prediction means 81 may use a past partial image, which is an image in which the portion containing the target object is extracted from a past image, as information indicating the target object, and predict the object presence region based on the correlation between the past partial image and the current image.

[0065] Specifically, the object presence region prediction means 81 may predict the object presence region based on multiple correlations calculated while sliding past partial images relative to the current image.

[0066] Alternatively, the object existence region prediction means 81 may use a deep learning model that takes two images as input and outputs the location with the highest correlation between the two images to predict the object existence region from the past partial image and the current image.

[0067] Some or all of the above embodiments may also be described as follows, but are not limited to the following:

[0068] (Appendix 1) An object existence region prediction means predicts the object existence region, which is the region in which the target object exists in the current image, based on information indicating the target object detected from past images, Object existence region fragment generation means for generating an object existence region fragment, which is a sub-region of the object existence region, from the object existence region, Object detection means for detecting an object detection fragment from the object presence region fragment which is a region containing the target object, The system includes object detection means for detecting the target object from the current image using the object detection fragment. An object detection system characterized by the following features.

[0069] (Note 2) The object presence region prediction means predicts the object presence region by using an object detection frame, which indicates the presence region of the target object detected from past images, as information indicating the target object. The target object detection means estimates an object detection frame indicating the region where the target object exists in the current image from the object detection frame and object detection fragment. The object detection system described in Appendix 1.

[0070] (Note 3) The target object detection means estimates the horizontal or vertical size of the object detection frame in the current image from the vertical and horizontal sizes of the object detection frame obtained from past images and the vertical and horizontal sizes of the detection frame obtained from object detection fragments. The object detection system described in Appendix 2.

[0071] (Note 4) The object existence region fragment generation means generates an object existence region fragment with the position of the object existence region fragment in the object existence region added to it. An object detection system described in any one of the appendices 1 through 3.

[0072] (Note 5) The object existence region fragment generation means generates an object existence region fragment with the position of the object existence region fragment relative to the object existence region before division added as the position of the object existence region fragment. The object detection system described in Appendix 4.

[0073] (Note 6) The object existence region fragment generation means generates object existence region fragments by dividing the object existence region into two equal parts in the vertical or horizontal direction. An object detection system described in any one of the appendices 1 through 3.

[0074] (Note 7) The object presence region prediction means uses a past partial image, which is an image in which the portion containing the target object is extracted from a past image, as information indicating the target object, and predicts the object presence region based on the correlation between the past partial image and the current image. The object detection system described in Appendix 1.

[0075] (Note 8) The object presence region prediction means predicts the object presence region based on multiple correlations calculated by sliding past partial images relative to the current image. The object detection system described in Appendix 7.

[0076] (Note 9) The object presence region prediction means uses a deep learning model that takes two images as input and outputs the location with the highest correlation between the two images to predict the object presence region from the past partial image and the current image. The object detection system described in Appendix 7.

[0077] (Note 10) The computer predicts the object existence region, which is the region in the current image where the object exists, based on information indicating the object detected from past images. The computer generates an object existence region fragment, which is a sub-region of the object existence region, from the object existence region. The computer detects an object detection fragment from the object existence region fragment that is the region containing the target object, The computer uses the object detection fragment to detect the target object from the current image. A method for detecting an object characterized by the following features.

[0078] (Note 11) To the computer, An object presence region prediction process that predicts the object presence region, which is the area in the current image where the target object is located, based on information indicating the target object detected from past images. A process for generating object existence region fragments, which are sub-regions of the object existence region, from the aforementioned object existence region. An object detection process that detects an object detection fragment from the object existence region fragment that is the region containing the target object, and Existence region detection process to detect the target object from the current image using the object detection fragment An object detection program to execute. [Industrial applicability]

[0079] The present invention is suitably applied to object detection systems that detect target objects in an image. For example, the present invention can be suitably applied to traffic systems that detect vehicles and people by object detection, and inspection systems that perform inspection by detecting products by object detection. [Explanation of symbols]

[0080] 100,110 Object Detection Systems 200,210 Object Location Predictor 300 Object existence region fragment generation unit 400 object detectors 500 Target Object Detectors 600 Current Images 610 Imaging device 620 Storage section 700 past images 710 Past partial image 800 Past Image Object Detection Results 900 Object detection results in the current image 1000 past partial image generator

Claims

1. An object presence region prediction means predicts the object presence region, which is the area in which the target object exists in the current image, based on information indicating the target object detected from past images. Object existence region fragment generation means for generating an object existence region fragment, which is a sub-region of the object existence region, from the object existence region, Object detection means for detecting an object detection fragment from the object presence region fragment which is a region containing the target object, The system includes object detection means for detecting the target object from the current image using the object detection fragment. An object detection system characterized by the following features.

2. The object presence region prediction means predicts the object presence region by using an object detection frame, which indicates the presence region of the target object detected from past images, as information indicating the target object. The target object detection means estimates an object detection frame indicating the region where the target object exists in the current image from the object detection frame and object detection fragment. The object detection system according to claim 1.

3. The target object detection means estimates the horizontal or vertical size of the object detection frame in the current image from the vertical and horizontal sizes of the object detection frame obtained from past images and the vertical and horizontal sizes of the detection frame obtained from object detection fragments. The object detection system according to claim 2.

4. The object existence region fragment generation means generates an object existence region fragment with the position of the object existence region fragment added to it within the object existence region. The object detection system according to any one of claims 1 to 3.

5. The object existence region fragment generation means generates an object existence region fragment with the position of the object existence region fragment added to its position relative to the object existence region before division. The object detection system according to claim 4.

6. The object existence region fragment generation means generates object existence region fragments by dividing the object existence region into two equal parts vertically or horizontally. The object detection system according to any one of claims 1 to 3.

7. The object presence region prediction means uses a past partial image, which is an image in which the portion containing the target object is extracted from a past image, as information indicating the target object, and predicts the object presence region based on the correlation between the past partial image and the current image. The object detection system according to claim 1.

8. The object presence region prediction means predicts the object presence region based on multiple correlations calculated by sliding past partial images relative to the current image. The object detection system according to claim 7.

9. The computer predicts the object existence region, which is the area in the current image where the target object exists, based on information indicating the target object detected from past images. The computer generates an object existence region fragment, which is a sub-region of the object existence region, from the object existence region. The computer detects an object detection fragment from the object existence region fragment that is the region containing the target object, The computer uses the object detection fragment to detect the target object from the current image. A method for detecting an object characterized by the following features.

10. On the computer, An object presence region prediction process that predicts the object presence region, which is the area in the current image where the target object is located, based on information indicating the target object detected from past images. A process for generating object existence region fragments, which are sub-regions of the object existence region, from the aforementioned object existence region. An object detection process that detects an object detection fragment from the object existence region fragment that is the region containing the target object, and Object detection process to detect the target object from the current image using the object detection fragment An object detection program to execute.

Images