Anomaly detection method, system, device, and storage medium

Abstract

The application discloses an abnormal object detection method, comprising the following steps: acquiring a plurality of reference images and determining a target region and establishing a background model based on the plurality of reference images; acquiring an image to be detected and predicting the frame image to be detected by using the background model to obtain a foreground image; determining a first region in the foreground image based on the coordinates of a plurality of target reference objects in the image to be detected, determining a second region in the foreground image based on the brightness difference and the chrominance difference between the image to be detected and the background model, and determining a third region in the foreground image based on the target region; and taking the residual image spot after removing the first region, the second region and the third region in the foreground image as an abnormal object in the target region. The application further discloses a system, a computer device and a readable storage medium. The application provides the scheme, which can effectively improve the detection efficiency and accuracy of the abnormal object.

Description

Technical Field

[0001] This invention relates to the field of object recognition, and specifically to an abnormal object detection method, system, device, and storage medium. Background Technology

[0002] Road obstructions refer to various unusual objects appearing on the road, such as spilled objects, construction sites, and warning barriers. Currently, road obstruction detection is usually carried out through manual monitoring or target-based detection methods. However, manual methods require a large investment of manpower and resources, resulting in low efficiency and high costs.

[0003] While object detection-based methods can identify anomalous objects in target areas, these methods require the collection and labeling of anomalous object samples. However, novel anomalous objects that have not been collected are difficult to detect accurately, leading to missed detections. Summary of the Invention

[0004] In view of this, in order to overcome at least one aspect of the above problems, embodiments of the present invention propose an anomaly detection method, comprising the following steps:

[0005] Acquire multiple reference images and determine the target region and establish a background model based on the multiple reference images;

[0006] The image to be detected is acquired, and the background model is used to predict the foreground image of the frame image to be detected.

[0007] A first region is determined in the foreground image based on the coordinates of several target reference objects in the image to be detected; a second region is determined in the foreground image based on the brightness and chromaticity differences between the image to be detected and the background model; and a third region is determined in the foreground image based on the target region.

[0008] The remaining patches after removing the first region, the second region, and the third region from the foreground image are considered as anomalies in the target region.

[0009] In some embodiments, determining the target region based on the multi-frame reference images further includes:

[0010] Semantic segmentation prediction is performed on each frame of the reference image to obtain the classification of each pixel in each frame of the reference image.

[0011] Statistically classify pixels at the same location in each frame of the reference image;

[0012] The target area is determined based on the statistical results.

[0013] In some embodiments, determining a first region in the foreground image based on the coordinates of several target reference objects in the image to be detected further includes:

[0014] The target detection model is used to detect the image to be detected to obtain the coordinates of multiple target reference objects, wherein the target reference objects include at least one of pedestrians, motor vehicles and non-motor vehicles;

[0015] Based on the coordinates of each target reference object, a corresponding first patch is determined in the foreground image, and it is determined whether each first patch has a connected second patch.

[0016] In response to the existence of the second patch, all the first patches and the second patch are collectively defined as the first region.

[0017] In some embodiments, determining a second region in the foreground image based on the brightness and chromaticity differences between the image to be detected and the background model further includes:

[0018] In the LUV color space, the image to be detected and the background model are compared, and the coordinates of the illuminated and shadow areas in the image to be detected with chromaticity differences less than a first threshold and brightness differences greater than a second threshold are determined.

[0019] The second region is determined in the foreground image based on the coordinates of the lighting and shadow areas.

[0020] In some embodiments, determining a third region in the foreground image based on the target region further includes:

[0021] The area outside the target area is designated as the third area.

[0022] In some embodiments, the remaining patch after removing the first region, the second region, and the third region from the foreground image is considered an anomaly in the target region, further comprising:

[0023] After removing the first region, the second region, and the third region, noise removal is performed using a data morphology algorithm.

[0024] The remaining patches in the foreground image are considered as anomalies in the target area.

[0025] In some embodiments, acquiring an image to be detected and using the background model to predict the foreground image of the frame image to be detected further includes:

[0026] The background model is updated using the image to be detected.

[0027] In some embodiments, acquiring multiple reference images and determining the target region and establishing a background model based on the multiple reference images further includes:

[0028] Use several frames from the multi-frame images as the background album;

[0029] The background album is divided into a first part, a second part, and a third part, and an update probability is set for each part, wherein the update probability of the first part is less than the update probability of the second part, and the update probability of the second part is less than the update probability of the third part.

[0030] The background album is updated using the image to be detected.

[0031] In some embodiments, the remaining patch after removing the first region, the second region, and the third region from the foreground image is considered an anomaly in the target region, further comprising:

[0032] Based on the coordinates of the anomaly, a slice of the anomaly is obtained in the image to be detected;

[0033] The name of the anomaly is obtained by identifying the slice of the anomaly.

[0034] In some embodiments, it also includes:

[0035] Save the slice, name, and coordinates of the anomaly to the database;

[0036] Determine whether the database records any anomalies with the same name and coordinates;

[0037] If the record exists, increment the count by 1.

[0038] If the record does not exist, the record count will be set to 1.

[0039] In some embodiments, it also includes:

[0040] In response to the detection of an anomaly being recorded more than a preset number of times, a slice of the corresponding position in the background album is obtained based on the coordinates of the anomaly;

[0041] A warning is generated in response to the fact that the similarity between a slice of a corresponding position in multiple frames of images in the background album and a slice of the anomaly is less than a threshold.

[0042] Based on the same inventive concept, according to another aspect of the present invention, embodiments of the present invention also provide an anomaly detection system, comprising:

[0043] The background modeling module is configured to acquire multiple frames of reference images and determine the target region and establish a background model based on the multiple frames of reference images;

[0044] The foreground prediction module is configured to acquire the image to be detected and use the background model to predict the foreground image of the frame image to be detected.

[0045] The filtering module is configured to determine a first region in the foreground image based on the coordinates of several target reference objects in the image to be detected, to determine a second region in the foreground image based on the brightness and chromaticity differences between the image to be detected and the background model, and to determine a third region in the foreground image based on the target region.

[0046] The detection module is configured to identify the remaining patches in the target area after removing the first region, the second region, and the third region from the foreground image.

[0047] Based on the same inventive concept, according to another aspect of the present invention, embodiments of the present invention also provide a computer device, comprising:

[0048] At least one processor; and

[0049] The memory stores a computer program that can run on the processor, which, when executing the program, performs the steps of any of the anomaly detection methods described above.

[0050] Based on the same inventive concept, according to another aspect of the present invention, embodiments of the present invention also provide a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of any of the anomaly detection methods described above.

[0051] The present invention has one of the following beneficial technical effects: The proposed solution provides anomaly detection in target areas based on background modeling, which has the advantages of high detection efficiency, high accuracy and low data acquisition cost. Furthermore, it reduces noise through various methods and can detect unknown objects without relying on a dataset, which can effectively improve the detection efficiency and accuracy of anomalies. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained based on these drawings without creative effort.

[0053] Figure 1A schematic flowchart of an anomaly detection method provided in an embodiment of the present invention;

[0054] Figure 2 A flowchart illustrating an anomaly detection method provided in an embodiment of the present invention;

[0055] Figure 3 A schematic diagram of the structure of an abnormal object detection system provided in an embodiment of the present invention;

[0056] Figure 4 A schematic diagram of the structure of a computer device provided for an embodiment of the present invention;

[0057] Figure 5 A schematic diagram of the structure of a computer-readable storage medium provided for an embodiment of the present invention. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to specific examples and the accompanying drawings.

[0059] It should be noted that all uses of "first" and "second" in the embodiments of the present invention are for the purpose of distinguishing two entities or parameters with the same name but different names. It is clear that "first" and "second" are only for the convenience of expression and should not be construed as limiting the embodiments of the present invention. Subsequent embodiments will not explain this in detail.

[0060] According to one aspect of the present invention, embodiments of the present invention provide an anomaly detection method, such as... Figure 1 As shown, it may include the following steps:

[0061] S1, acquire multiple frames of reference images and determine the target region and establish a background model based on the multiple frames of reference images;

[0062] S2, acquire the image to be detected and use the background model to predict the foreground image of the frame image to be detected;

[0063] S3, a first region is determined in the foreground image based on the coordinates of several target reference objects in the image to be detected; a second region is determined in the foreground image based on the brightness and chromaticity differences between the image to be detected and the background model; and a third region is determined in the foreground image based on the target region.

[0064] S4, the remaining patches after removing the first region, the second region and the third region in the foreground image are regarded as anomalies in the target region.

[0065] This invention provides a proposed scheme for anomaly detection in target areas based on background modeling, which has the advantages of high detection efficiency, high accuracy, and low data acquisition cost. Furthermore, it reduces noise through various methods and can detect unknown objects without relying on a dataset, thus effectively improving the detection efficiency and accuracy of anomalies.

[0066] In some embodiments, the solution proposed in this invention can be applied to road anomaly detection. Through steps S1-S4, anomalies in the road can be detected and identified, thereby ensuring road traffic safety.

[0067] In some embodiments, S1, determining the target region based on the multi-frame reference images further includes:

[0068] Semantic segmentation prediction is performed on each frame of the image to obtain the classification of each pixel in each frame of the image;

[0069] Statistically classify pixels at the same location in each frame of the image;

[0070] The target area is determined based on the statistical results.

[0071] Specifically, such as Figure 2 As shown, N frames of image data can be collected, and then a target scene semantic segmentation model trained on a manually labeled dataset can be used to perform semantic segmentation prediction, thereby obtaining the classification of each pixel. After performing semantic segmentation prediction on N frames of images, there are N classification results for pixels at the same location. The N classification results are statistically analyzed. If the proportion of classification results that are target regions is greater than a threshold, for example, if more than 60% of the N classification results for a pixel are target regions, then the final classification result of that pixel is considered to be the target region, and thus the target region can be obtained.

[0072] It should be noted that in this invention, the reference image and the image to be detected are of the same size in each frame.

[0073] In some embodiments, in step S1, such as Figure 2 As shown, when building a background model based on multiple reference images, the background model (BM) can be obtained by first using an illumination compensation algorithm to reduce the influence of illumination on N frames of images, using a Gaussian blur algorithm to remove image noise, and finally using a Gaussian mixture model (GMM) to model the background of the preprocessed N frames of images.

[0074] In some embodiments, S3, determining a first region in the foreground image based on the coordinates of several target reference objects in the image to be detected further includes:

[0075] The target detection model is used to detect the image to be detected to obtain the coordinates of multiple target reference objects, wherein the target reference objects include at least one of pedestrians, motor vehicles and non-motor vehicles;

[0076] Based on the coordinates of each target reference object, a corresponding first patch is determined in the foreground image, and it is determined whether each first patch has a connected second patch.

[0077] In response to the existence of the second patch, all the first patches and the second patch are collectively defined as the first region.

[0078] Specifically, such as Figure 2 As shown, an image to be detected can be acquired using a camera, and then the target detection model can be used to detect the image to obtain the coordinates of multiple target reference objects. In a road scene, the target reference objects can include at least one of pedestrians, motor vehicles, and non-motor vehicles. Then, a background model is used to predict the foreground image (FG) of the image to be detected. Based on the coordinates of the target reference objects, it is possible to determine which patches in the foreground image belong to the target reference objects. Since the target reference objects are not considered anomalies, they are considered interference items in the foreground image and need to be removed. Furthermore, since the target reference objects have shadows, the shadow patches also need to be removed. Generally speaking, the shadow and the target reference object are connected. In the foreground image, this means that the first patch of the target reference object and the second patch of the shadow are connected. Therefore, all the first patches and second patches are considered as the first region F1 to be removed.

[0079] In some embodiments, S3, determining a second region in the foreground image based on the brightness and chromaticity differences between the image to be detected and the background model further includes:

[0080] In the LUV color space, the image to be detected and the background model are compared, and the coordinates of the illuminated and shadow areas in the image to be detected with chromaticity differences less than a first threshold and brightness differences greater than a second threshold are determined.

[0081] The second region is determined in the foreground image based on the coordinates of the lighting and shadow areas.

[0082] Specifically, such as Figure 2 As shown, the RGB color space of the image to be detected is converted to the LUV color space by color space transformation, and then the L value is equalized to obtain the light and shadow areas where the color of the image to be detected and the background model are similar but the brightness is different. Based on the coordinates of the light and shadow areas, the second region F2 that needs to be removed in the foreground image can be determined.

[0083] In some embodiments, S3, determining a third region in the foreground image based on the target region further includes:

[0084] The area outside the target area is designated as the third area.

[0085] Specifically, since the anomalies that need to be detected are in the target area, the area outside the target area is designated as the third area F3 that needs to be removed.

[0086] In some embodiments, the remaining patch after removing the first region, the second region, and the third region from the foreground image is considered an anomaly in the target region, further comprising:

[0087] After removing the first region, the second region, and the third region, noise removal is performed using a data morphology algorithm.

[0088] The remaining patches in the foreground image are considered as anomalies in the target area.

[0089] Specifically, after removing the first, second, and third regions from the foreground image, mathematical morphology algorithms (such as erosion and dilation algorithms) can be used to remove small spots to reduce noise. The remaining foreground spots are the suspected abnormalities after screening and filtering.

[0090] In some embodiments, acquiring an image to be detected and using the background model to predict the foreground image of the frame image to be detected further includes:

[0091] The background model is updated using the image to be detected.

[0092] In some embodiments, acquiring multiple reference images and determining the target region and establishing a background model based on the multiple reference images further includes:

[0093] Use several frames from the multi-frame images as the background album;

[0094] The background album is divided into a first part, a second part, and a third part, and an update probability is set for each part, wherein the update probability of the first part is less than the update probability of the second part, and the update probability of the second part is less than the update probability of the third part.

[0095] The background album is updated using the image to be detected.

[0096] Specifically, a portion of N reference images is randomly stored as a background album (BH), and the background album can be updated using the image to be detected. In some embodiments, the update operation of the BH is probabilistic. Based on this, the BH is divided into three segments, A, B, and C, according to its length. The background album of segment A has a very low probability of being updated, the update probability of segment B is relatively higher than that of segment A, and the update probability of segment C is the highest, thereby enabling the BH to record the longest possible background history.

[0097] In some embodiments, the remaining patch after removing the first region, the second region, and the third region from the foreground image is considered an anomaly in the target region, further comprising:

[0098] Based on the coordinates of the anomaly, a slice of the anomaly is obtained in the image to be detected;

[0099] The name of the anomaly is obtained by identifying the slice of the anomaly.

[0100] In some embodiments, it also includes:

[0101] Save the slice, name, and coordinates of the anomaly to the database;

[0102] Determine whether the database records any anomalies with the same name and coordinates;

[0103] If the record exists, increment the count by 1.

[0104] If the record does not exist, the record count will be set to 1.

[0105] In some embodiments, it also includes:

[0106] In response to the detection of an anomaly being recorded more than a preset number of times, a slice of the corresponding position in the background album is obtained based on the coordinates of the anomaly;

[0107] A warning is generated in response to the fact that the similarity between a slice of a corresponding position in multiple frames of images in the background album and a slice of the anomaly is less than a threshold.

[0108] Specifically, all suspected anomalies can be assigned a record count of 1 (initial value is 1), and the slice, name and coordinates of the suspected anomaly are saved in the database. If the name and coordinates of the suspected anomaly detected in subsequent consecutive frames in the database are the same as the name and coordinates of the suspected anomaly detected in this detection, the record count of the suspected anomaly in the database is incremented by 1. If they are different, it will be saved as a new suspected anomaly in the database and assigned a record count of 1.

[0109] Once the number of records of suspected anomalies reaches a certain number, it indicates that the object has been present in a certain location for a sufficient period of time. An image similarity algorithm is then used to compare the similarity between the object and slices at the same location in the image frame (BH). If the similarity is low across BH images of a certain length, the object can be considered an anomaly. This prevents misjudgment of anomalies based on a single image.

[0110] The detection report of an anomaly can be used to issue an early warning or be sent to vehicles on the road to provide real-time road condition alerts. Simultaneously, the detection results are stored in a database for processing and used to create an anomaly sample set, further improving algorithm performance. Furthermore, the anomaly can be tracked, and the warning can be lifted once the anomaly is identified and cleared.

[0111] This invention provides a proposed scheme for anomaly detection in target areas based on background modeling, which has the advantages of high detection efficiency, high accuracy, and low data acquisition cost. Furthermore, it reduces noise through various methods and can detect unknown objects without relying on a dataset, thus effectively improving the detection efficiency and accuracy of anomalies.

[0112] Based on the same inventive concept, according to another aspect of the present invention, embodiments of the present invention also provide an anomaly detection system 400, such as... Figure 3 As shown, it includes:

[0113] Background modeling module 401 is configured to acquire multiple frames of reference images and determine the target region and establish a background model based on the multiple frames of reference images;

[0114] The foreground prediction module 402 is configured to acquire the image to be detected and use the background model to predict the foreground image of the frame image to be detected to obtain a foreground image;

[0115] The filtering module 403 is configured to determine a first region in the foreground image based on the coordinates of several target reference objects in the image to be detected, to determine a second region in the foreground image based on the brightness and chromaticity differences between the image to be detected and the background model, and to determine a third region in the foreground image based on the target region.

[0116] The detection module 404 is configured to identify the remaining patches after removing the first region, the second region, and the third region from the foreground image as anomalies in the target region.

[0117] Based on the same inventive concept, according to another aspect of the present invention, such as Figure 4 As shown, embodiments of the present invention also provide a computer device 501, comprising:

[0118] At least one processor 520; and

[0119] The memory 510 stores a computer program 511 that can run on the processor. When the processor 520 executes the program, it performs the steps of any of the above-described anomaly detection methods.

[0120] Based on the same inventive concept, according to another aspect of the present invention, such as Figure 5 As shown, embodiments of the present invention also provide a computer-readable storage medium 601, which stores a computer program 610. When the computer program 610 is executed by a processor, it performs the steps of any of the above-described anomaly detection methods.

[0121] Finally, it should be noted that those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods.

[0122] Furthermore, it should be understood that the computer-readable storage medium (e.g., memory) described herein may be volatile memory or non-volatile memory, or may include both volatile memory and non-volatile memory.

[0123] Those skilled in the art will also understand that the various exemplary logic blocks, modules, circuits, and algorithm steps described in conjunction with the disclosure herein can be implemented as electronic hardware, computer software, or a combination of both. To clearly illustrate this interchangeability between hardware and software, the functionality of various illustrative components, blocks, modules, circuits, and steps has been generally described. Whether this functionality is implemented as software or as hardware depends on the specific application and the design constraints imposed on the system as a whole. Those skilled in the art can implement the functionality in various ways for each specific application, but such implementation decisions should not be construed as departing from the scope of the embodiments disclosed herein.

[0124] The above are exemplary embodiments disclosed in this invention. However, it should be noted that various changes and modifications can be made without departing from the scope of the embodiments of this invention as defined by the claims. The functions, steps, and / or actions of the methods according to the disclosed embodiments described herein do not need to be performed in any particular order. Furthermore, although the elements disclosed in the embodiments of this invention may be described or claimed individually, they may be understood as multiple unless explicitly limited to a singular number.

[0125] It should be understood that, as used herein, the singular form “a” is intended to include the plural form as well, unless the context clearly supports an exception. It should also be understood that, as used herein, “and / or” refers to any and all possible combinations of one or more of the associated listed items.

[0126] The embodiment numbers disclosed in the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0127] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0128] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of the invention (including the claims) is limited to these examples. Within the framework of the invention, technical features of the above embodiments or different embodiments can be combined, and many other variations of different aspects of the invention exist, which are not provided in the details for the sake of brevity. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the invention should be included within the protection scope of the invention.

Claims

1. A method for detecting abnormal objects, characterized in that, Includes the following steps: Acquire multiple reference images and determine the target region and establish a background model based on the multiple reference images; The image to be detected is acquired, and the background model is used to predict the foreground image of the frame image to be detected. A first region is determined in the foreground image based on the coordinates of several target reference objects in the image to be detected; a second region is determined in the foreground image based on the brightness and chromaticity differences between the image to be detected and the background model; and the region outside the target region is taken as a third region. The remaining patches after removing the first region, the second region, and the third region from the foreground image are considered as anomalies in the target region. Determining a second region in the foreground image based on the brightness and chromaticity differences between the image to be detected and the background model further includes: In the LUV color space, the image to be detected and the background model are compared, and the coordinates of the illuminated and shadow areas in the image to be detected with chromaticity differences less than a first threshold and brightness differences greater than a second threshold are determined. A second region is determined in the foreground image based on the coordinates of the lighting and shadow regions; The establishment of the background model further includes: Create a background album, which is divided into a first part, a second part, and a third part according to its length, and each part has a different update probability.

2. The method as described in claim 1, characterized in that, Determining the target region based on the multi-frame reference images further includes: Semantic segmentation prediction is performed on each frame of the reference image to obtain the classification of each pixel in each frame of the reference image. Statistically classify pixels at the same location in each frame of the reference image; The target area is determined based on the statistical results.

3. The method as described in claim 1, characterized in that, Determining a first region in the foreground image based on the coordinates of several target reference objects in the image to be detected further includes: The target detection model is used to detect the image to be detected to obtain the coordinates of multiple target reference objects, wherein the target reference objects include at least one of pedestrians, motor vehicles and non-motor vehicles; Based on the coordinates of each target reference object, a corresponding first patch is determined in the foreground image, and it is determined whether each first patch has a connected second patch. In response to the existence of the second patch, all the first patches and the second patch are collectively defined as the first region.

4. The method as described in claim 1, characterized in that, The remaining patch after removing the first region, the second region, and the third region from the foreground image is considered an anomaly in the target region, further comprising: After removing the first region, the second region, and the third region, noise removal is performed using a data morphology algorithm. The remaining patches in the foreground image are considered as anomalies in the target area.

5. The method as described in claim 1, characterized in that, Acquiring the image to be detected and using the background model to predict the foreground image of the frame image to be detected further includes: The background model is updated using the image to be detected.

6. The method as described in claim 1, characterized in that, Acquiring multiple reference images and determining the target region and establishing a background model based on the multiple reference images further includes: Use several frames from the multi-frame images as the background album; Set an update probability for each part, wherein the update probability of the first part is less than the update probability of the second part, and the update probability of the second part is less than the update probability of the third part. The background album is updated using the image to be detected.

7. The method as described in claim 6, characterized in that, The remaining patch after removing the first region, the second region, and the third region from the foreground image is considered an anomaly in the target region, further comprising: Based on the coordinates of the anomaly, a slice of the anomaly is obtained in the image to be detected; The name of the anomaly is obtained by identifying the slice of the anomaly.

8. The method as described in claim 7, characterized in that, Also includes: Save the slice, name, and coordinates of the anomaly to the database; Determine whether the database records any anomalies with the same name and coordinates; If the record exists, increment the count by 1. If the record does not exist, the record count will be set to 1.

9. The method as described in claim 8, characterized in that, Also includes: In response to the detection of an anomaly being recorded more than a preset number of times, a slice of the corresponding position in the background album is obtained based on the coordinates of the anomaly; A warning is generated in response to the fact that the similarity between a slice of a corresponding position in multiple frames of images in the background album and a slice of the anomaly is less than a threshold.

10. An abnormal object detection system, characterized in that, include: The background modeling module is configured to acquire multiple frames of reference images and determine the target region and establish a background model based on the multiple frames of reference images; The foreground prediction module is configured to acquire the image to be detected and use the background model to predict the foreground image of the frame image to be detected. The filtering module is configured to determine a first region in the foreground image based on the coordinates of several target reference objects in the image to be detected, determine a second region in the foreground image based on the brightness and chromaticity differences between the image to be detected and the background model, and designate the region outside the target region as a third region. The detection module is configured to identify the remaining patches after removing the first region, the second region, and the third region from the foreground image as anomalies in the target region. The filtering module is also configured as follows: In the LUV color space, the image to be detected and the background model are compared, and the coordinates of the illuminated and shadow areas in the image to be detected with chromaticity differences less than a first threshold and brightness differences greater than a second threshold are determined. A second region is determined in the foreground image based on the coordinates of the lighting and shadow regions; The background modeling module is also configured as follows: Create a background album, which is divided into a first part, a second part, and a third part according to its length, and each part has a different update probability.

11. A computer device, comprising: At least one processor; as well as A memory storing a computer program executable on the processor, characterized in that the processor executes the program by performing the steps of the method as described in any one of claims 1-9.

12. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it performs the steps of the method as described in any one of claims 1-9.

Images