Shadow detection method, system, device and medium based on BSD perspective

Abstract

The application provides a BSD view-based shadow detection method, system, device and medium, and relates to the technical field of vehicle auxiliary driving. The method comprises the following steps: in the starting stage of the vehicle, BSD images captured by a BSD camera of the vehicle in a current period are acquired; a first segmentation model is used to perform semantic segmentation on the shadow area of the BSD images, so as to obtain a plurality of frame shadow area segmentation results in the current period; after the vehicle enters a normal driving state, the BSD images acquired in real time and the plurality of frame shadow area segmentation results are input into a second segmentation model, and a segmentation result change amount of a current frame relative to a previous frame is obtained through the second segmentation model; the shadow area segmentation result of the previous frame obtained from the plurality of frame shadow area segmentation results is superimposed with the segmentation result change amount, so as to obtain a shadow area detection result of the current frame. The application can effectively improve the shadow detection accuracy of the BSD system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of vehicle auxiliary driving, in particular to a shadow detection method, system, device and medium based on BSD view. BACKGROUND

[0002] The automobile BSD (blind spot detection) is a vehicle blind area monitoring system, which is used to detect whether there is a person or vehicle in the rear of the adjacent lane and the blind area of the rearview mirror. When a person or vehicle approaches or has entered the blind area, the monitoring system will remind the driver through sound, light and other ways. One of the implementation ways of BSD is to use a camera to cooperate with a target detection model to detect a person or vehicle in a certain range, but when a shadow area appears in the BSD view, it may interfere with the detection of the person or vehicle, so the shadow detection function of the BSD system is also very important.

[0003] At present, the existing scheme mainly uses an AI segmentation model to perform semantic segmentation on a single frame picture of the BSD view, so as to determine the position of the shadow area. Due to the computing power limitation of the vehicle-mounted chip, the BSD system usually adopts a lightweight model, and the segmentation effect and accuracy are not ideal; secondly, because the model usually only receives a single frame picture when working, the continuity of the image during vehicle driving is not considered, and the result of the shadow segmentation may also jump between frames, thereby affecting the detection accuracy of the shadow area. SUMMARY

[0004] Therefore, the embodiments of the present application provide a shadow detection method, system, device and medium based on BSD view, so as to improve the shadow detection accuracy of the BSD system.

[0005] The embodiments of the present application provide the following technical solutions: a shadow detection method based on BSD view, comprising:

[0006] In the starting stage of the vehicle, BSD images captured by a vehicle-mounted BSD camera in a current period are acquired;

[0007] A first segmentation model is used to perform semantic segmentation on the shadow area of the BSD image, and a plurality of frame shadow area segmentation results in the current period are obtained;

[0008] After the vehicle enters a normal driving state, the BSD image acquired in real time and the plurality of frame shadow area segmentation results are input into a second segmentation model, and a segmentation result change amount of a current frame relative to a previous frame is obtained by prediction of the second segmentation model;

[0009] The shadow area segmentation result of the previous frame obtained from the plurality of frame shadow area segmentation results is superimposed with the segmentation result change amount, and a shadow area detection result of the current frame is obtained.

[0010] According to an embodiment of the present application, the first segmentation model is used to perform semantic segmentation on the shadow area of the preprocessed BSD image, to obtain a plurality of shadow area segmentation results in a current time period, including:

[0011] The first segmentation model is used to perform semantic segmentation on the shadow area of the preprocessed BSD image, to obtain shadow area segmentation results of at least the first 10 frames in the current time period.

[0012] The shadow area segmentation results of the at least first 10 frames are saved into a continuous frame segmentation result pool.

[0013] According to an embodiment of the present application, the shadow area detection result of the current frame is obtained, further including:

[0014] The shadow area detection result of the current frame is saved into the continuous frame segmentation result pool, as a shadow area segmentation result of a next frame input into the second segmentation model.

[0015] According to an embodiment of the present application, the first segmentation model is used to perform semantic segmentation on the shadow area of the preprocessed BSD image, to obtain a plurality of shadow area segmentation results in a current time period, further including:

[0016] The shadow area segmentation result is represented as a matrix of element 0 and element 1, element 0 indicating that the area in the BSD image is not a shadow, and element 1 indicating that the area in the BSD image is a shadow.

[0017] According to an embodiment of the present application, the second segmentation model is used to predict a segmentation result change amount of the current frame relative to a previous frame, including:

[0018] The segmentation result change amount is represented as a matrix of element True and element False, element True indicating that the element of the area in the BSD image in the current frame is consistent with the element of the area in the previous frame, and element False indicating that the element of the area in the BSD image in the current frame is inconsistent with the element of the area in the previous frame.

[0019] According to an embodiment of the present application, the shadow area segmentation result of the previous frame obtained from the plurality of shadow area segmentation results is superimposed with the segmentation result change amount, to obtain a shadow area detection result of the current frame, including:

[0020] According to the element value of the matrix of the segmentation result change amount, the shadow area segmentation result of the previous frame is modified.

[0021] If the element value of the matrix of the segmentation result change amount is True, the shadow area segmentation result of the previous frame is the shadow area detection result of the current frame.

[0022] If the element value of the matrix of the segmentation result change amount is False, the shadow region segmentation result of the previous frame is flipped to obtain the shadow region detection result of the current frame.

[0023] According to an embodiment of the present application, during the starting stage of the vehicle, the BSD image captured by the BSD camera of the vehicle in the current period is obtained, and the method further comprises:

[0024] The BSD image is preprocessed.

[0025] The preprocessing process comprises image scaling, image color mapping and image normalization.

[0026] The present application also provides a shadow detection system based on a BSD perspective, comprising:

[0027] An image acquisition module is configured to obtain a BSD image captured by a BSD camera of a vehicle in a current period during a starting stage of the vehicle.

[0028] A first model processing module is configured to perform semantic segmentation on a shadow region of the BSD image by using a first segmentation model to obtain a plurality of frames of shadow region segmentation results in the current period.

[0029] A second model processing module is configured to input the BSD image acquired in real time and the plurality of frames of shadow region segmentation results into a second segmentation model after the vehicle enters a normal driving state, and to obtain a segmentation result change amount of a current frame relative to a previous frame by using the second segmentation model for prediction.

[0030] A detection result output module is configured to superimpose a shadow region segmentation result of a previous frame obtained from the plurality of frames of shadow region segmentation results and the segmentation result change amount to obtain a shadow region detection result of a current frame.

[0031] The present application also provides a computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the shadow detection method based on a BSD perspective as described above when executing the computer program.

[0032] The present application also provides a computer readable storage medium storing a computer program for executing the shadow detection method based on a BSD perspective as described above.

[0033] Compared with the prior art, the at least one technical solution adopted by the embodiment of the present specification can achieve the beneficial effects at least including: compared with the single-frame shadow segmentation model scheme, the embodiment of the present application significantly improves the segmentation accuracy, adopts two models of large and small, and the computing power requirement is also significantly reduced, thereby meeting the needs of BSD system shadow segmentation. Compared with the single-frame scheme, by inputting the segmentation results of consecutive frames, the continuity between frames is considered, and the segmentation result is more unified and coherent; secondly, the task of the small model is simplified from predicting the shadow area to predicting the change amount of the last frame shadow segmentation result, effectively reducing the task difficulty, and the computing power requirement of the model is also reduced, on the basis of improving the shadow detection accuracy of the BSD system, further ensuring the real-time performance of the BSD system. BRIEF DESCRIPTION OF DRAWINGS

[0034] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0035] Figure 1 is a BSD view-based shadow detection method schematic diagram of the embodiment of the present application;

[0036] Figure 2 is a shadow detection method flowchart schematic diagram of the embodiment of the present application;

[0037] Figure 3 is a BSD view-based shadow detection system structure block diagram of the embodiment of the present application;

[0038] Figure 4 is a structure schematic diagram of the computer device of the present application. DETAILED DESCRIPTION

[0039] The embodiments of the present application will be described in detail below with reference to the drawings.

[0040] Following make the application's implementation through specific concrete example, the person skilled in the art can easily understand the other advantages and effects of the application from the disclosure of the present application. Obviously, the described embodiments are only a part of the embodiments of the application, not all the embodiments. The application can also be implemented or applied by another different specific implementation, and the details in the specification can be based on different views and applications, various modifications or changes are made without departing from the spirit of the application. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict. Based on the embodiments in the present application, all other embodiments obtained by the person skilled in the art without creative labor belong to the scope of protection of the present application.

[0041] As shown in Figure 1 The embodiment of the application provides a shadow detection method based on BSD view, comprising:

[0042] S101. In the starting stage of the vehicle, the BSD image captured by the BSD camera of the vehicle in the current period is acquired;

[0043] S102. The shadow area of the BSD image is segmented by using a first segmentation model, and a plurality of frame shadow area segmentation results in the current period are obtained;

[0044] S103. After the vehicle enters the normal driving state, the BSD image acquired in real time and the plurality of frame shadow area segmentation results are input into a second segmentation model, and the segmentation result change amount of the current frame relative to the previous frame is obtained by predicting the second segmentation model;

[0045] S104. The shadow area segmentation result of the previous frame obtained from the plurality of frame shadow area segmentation results is superimposed with the segmentation result change amount, and the shadow area detection result of the current frame is obtained.

[0046] In the embodiment, the input of the first segmentation model is the BSD image, and the output is the shadow area segmentation result. The input of the second segmentation model is the real-time BSD image and the shadow area segmentation result output by the first segmentation model, and the output is the segmentation result change amount of the current frame relative to the previous frame. The segmentation result change amount of the current frame relative to the previous frame is superimposed with the shadow segmentation result of the previous frame, and the shadow area segmentation result of the current frame is obtained.

[0047] In a specific implementation, the first segmentation model is a large model, which refers to a deep neural network model with millions or billions of parameters. The large model has a large number of parameters and can perform complex processing and task processing on large-scale data. The second segmentation model is a small model, which refers to a deep neural network model with fewer parameters. The small model has advantages such as lightweight, high efficiency, and easy deployment. Common small models include linear models, decision trees, and naive Bayes, etc. These models are suitable for scenarios with small amounts of data and limited computing resources, such as mobile applications, embedded devices, and Internet of Things, etc.

[0048] The embodiment of the present application improves the detection capability of the BSD system for shadows by deploying a large and small AI model in the BSD system to detect shadow areas, on the one hand, the large model only performs accurate detection of shadows during the starting stage of the vehicle, on the other hand, the subsequent small model only needs to predict the change value of the current frame and the previous frame shadow, thereby reducing the computational amount and improving the real-time performance of the system.

[0049] In one embodiment, a first segmentation model is used to perform semantic segmentation on the shadow area of the preprocessed BSD image, and a plurality of frame shadow area segmentation results in a current period are obtained, including:

[0050] The first segmentation model is used to perform semantic segmentation on the shadow area of the preprocessed BSD image, and at least the first 10 frame shadow area segmentation results in the current period are obtained;

[0051] The at least first 10 frame shadow area segmentation results are saved to a continuous frame segmentation result pool.

[0052] Compared with the model input scheme of a single frame picture, the embodiment considers the continuity between frames by inputting the segmentation results of continuous frames, and the segmentation results are more unified and coherent. Secondly, the task of the small model is simplified from predicting the shadow area to predicting the change amount of the previous frame shadow segmentation result, effectively reducing the task difficulty, and the computing power requirement of the model is also reduced. On the basis of improving the shadow detection accuracy of the BSD system, the real-time performance of the BSD system is further ensured.

[0053] In a specific implementation, obtaining the shadow area detection result of the current frame further includes:

[0054] The shadow area detection result of the current frame is saved to the continuous frame segmentation result pool as the shadow area segmentation result of the next frame input into the second segmentation model. In the embodiment, the output result of the second segmentation model, i.e., the shadow area detection result of the current frame, is updated to the continuous frame segmentation result pool synchronously as the input of the next frame model, which ensures the real-time performance of the system detection.

[0055] As Figure 2 shown, the embodiment takes a certain BSD camera as an example for illustration, and other cameras perform similar operations. In the starting stage of the vehicle, the received BSD image is subjected to semantic segmentation of the shadow area, and is saved to a continuous frame segmentation result pool. After 10 frames of accurate segmentation are completed, in order to ensure the real-time performance of the BSD system during vehicle driving, a small model is started to be called for prediction. The small model receives two-way input, the first-way input being an image in the BSD camera, and the second-way input being the segmentation result of the previous 10 frames in the continuous frame segmentation result pool, and the output being a segmentation result change amount of the current frame relative to the previous frame, and then the segmentation result of the previous frame is superimposed to obtain the shadow area detection result of the current frame.

[0056] In specific implementation, the first segmentation model is used to perform semantic segmentation on the shadow area of the preprocessed BSD image to obtain a plurality of frames of shadow area segmentation results in a current period, and further includes:

[0057] The shadow area segmentation result is represented as a matrix of element 0 and element 1, element 0 indicating that the area in the BSD image is not a shadow, and element 1 indicating that the area in the BSD image is a shadow.

[0058] The segmentation result change amount of the current frame relative to the previous frame is obtained by prediction of the second segmentation model, and includes:

[0059] The segmentation result change amount is represented as a matrix of element True and element False, element True indicating that the element of the area in the BSD image in the current frame is consistent with the element of the area in the previous frame, and element False indicating that the element of the area in the BSD image in the current frame is inconsistent with the element of the area in the previous frame.

[0060] The shadow area segmentation result of the previous frame obtained from the plurality of frames of shadow area segmentation results is superimposed with the segmentation result change amount to obtain a shadow area detection result of the current frame, and includes:

[0061] According to the element value of the matrix of the segmentation result change amount, the shadow area segmentation result of the previous frame is modified;

[0062] If the element value of the matrix of the segmentation result change amount is True, the shadow area segmentation result of the previous frame is the shadow area detection result of the current frame;

[0063] If the element value of the matrix of the segmentation result change amount is False, the shadow area segmentation result of the previous frame after being flipped is the shadow area detection result of the current frame.

[0064] In specific implementation, since the shadow area segmentation result is represented by two results of element "0" and element "1", if the element of the current frame of the area in the BSD image is consistent with the element of the last frame of the area, the current frame remains the "0" or "1" of the element of the last frame, if the element of the current frame of the area in the BSD image is inconsistent with the element of the last frame of the area, the segmentation result of the last frame of the area is flipped to obtain the shadow area segmentation result of the current frame, that is, the element "0" is flipped to the element "1", or the element "1" is flipped to the element "0".

[0065] In specific implementation, in the starting stage of the vehicle, the BSD image captured by the BSD camera of the vehicle in the current period is obtained, and the BSD image further comprises:

[0066] The BSD image is preprocessed.

[0067] The preprocessing process comprises: (1) image scaling, scaling the image size from 1280x720 to 576x320; (2) image color mapping, mapping the image color from BGR to RGB; (3) image normalization, normalizing the image from the pixel value distribution [0, 255] to [0, 1].

[0068] In the embodiment of the application, two AI segmentation models, a large model and a small model, are used, the large model only works in the starting stage of the vehicle, and accurately segments the shadow area of the first 10 frames, and subsequently, in order to ensure the real-time performance of semantic segmentation, the small model is called to work, at this time, the output of the small model is no longer the shadow area of the current frame, but the change amount of the current frame relative to the segmentation result of the previous frame, and the shadow segmentation result of the previous frame is superimposed to obtain the shadow prediction area of the current frame. The large model can effectively ensure the accuracy of the shadow area segmentation in the starting stage of the vehicle, the small model can ensure the real-time performance of the system, and the method of predicting the change amount effectively reduces the task difficulty and considers the connection between frames.

[0069] As shown in Figure 3 The application further provides a shadow detection system 200 based on a BSD view, comprising:

[0070] An image acquisition module 201 is configured to acquire, in the starting stage of the vehicle, a BSD image captured by a BSD camera of the vehicle in a current period;

[0071] A first model processing module 202 is configured to perform semantic segmentation on a shadow area of the BSD image by using a first segmentation model to obtain a plurality of frame shadow area segmentation results in the current period.

[0072] The second model processing module 203 is configured to input the real-time acquired BSD image and the multiple frames of shadow region segmentation results into a second segmentation model after the vehicle enters the normal driving state, and obtain a segmentation result change amount of a current frame relative to a previous frame by using the second segmentation model.

[0073] The detection result output module 204 is configured to superimpose the shadow region segmentation result of the previous frame obtained from the multiple frames of shadow region segmentation results and the segmentation result change amount to obtain a shadow region detection result of a current frame.

[0074] In a specific implementation, the first model processing module 202 is configured to perform semantic segmentation on the shadow region of the preprocessed BSD image by using a first segmentation model to obtain shadow region segmentation results of at least the previous 10 frames in a current period; and save the shadow region segmentation results of the at least previous 10 frames into a continuous frame segmentation result pool.

[0075] In a specific implementation, the detection result output module 204 is further configured to save the shadow region detection result of the current frame into the continuous frame segmentation result pool as a shadow region segmentation result of a next frame input into the second segmentation model.

[0076] In a specific implementation, the first model processing module 202 is configured to represent the shadow region segmentation result as a matrix of element 0 and element 1, where element 0 represents that the region in the BSD image is not a shadow, and element 1 represents that the region in the BSD image is a shadow.

[0077] In a specific implementation, the second model processing module 203 is configured to represent the segmentation result change amount as a matrix of element True and element False, where element True represents that the element of the region in the BSD image in a current frame is consistent with the element of the region in a previous frame, and element False represents that the element of the region in the BSD image in the current frame is inconsistent with the element of the region in the previous frame.

[0078] In a specific implementation, the detection result output module 204 is configured to modify the shadow region segmentation result of the previous frame according to the element value of the matrix of the segmentation result change amount; if the element value of the matrix of the segmentation result change amount is True, the shadow region segmentation result of the previous frame is the shadow region detection result of the current frame; if the element value of the matrix of the segmentation result change amount is False, the shadow region segmentation result of the previous frame after being flipped is the shadow region detection result of the current frame.

[0079] In a specific implementation, the image acquisition module 201 is further configured to perform preprocessing on the BSD image captured by the BSD camera on board after acquiring the BSD image in the current period.

[0080] In one embodiment, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable in the processor, wherein the processor implements any of the BSD view based shadow detection methods described above when executing the computer program. Figure 4 As shown, the computer device comprises a memory 301, a processor 302, and a computer program stored in the memory and executable in the processor, wherein the processor implements any of the BSD view based shadow detection methods described above when executing the computer program.

[0081] In particular, the computer device can be a computer terminal, a server, or similar computing device.

[0082] In this embodiment, a computer readable storage medium is provided, which stores a computer program for implementing any of the BSD view based shadow detection methods described above.

[0083] In particular, the computer readable storage medium includes permanent and non-permanent, removable and non-removable media, which can be implemented by any method or technology to store information. The information can be computer readable instructions, data structures, program modules or other data. Examples of computer readable storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read only memory (ROM), electrically erasable programmable read only memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape disk storage or other magnetic storage device, or any other non-transmission medium that can be used to store information accessible by a computing device. According to the definition herein, computer readable storage media does not include transitory media such as modulated data signals and carriers.

[0084] Obviously, those skilled in the art should understand that each module or step of the above-mentioned embodiments of the present application can be realized by a general computing device, which can be concentrated on a single computing device or distributed on a network composed of multiple computing devices, and optionally, they can be realized by program codes executable by a computing device, so that they can be stored in a storage device and executed by a computing device, and in some cases, the steps shown or described herein can be executed in different order, or they can be manufactured into individual integrated circuit modules, or multiple modules or steps thereof can be manufactured into a single integrated circuit module. Thus, the embodiments of the present application are not limited to any specific combination of hardware and software.

[0085] The above merely provides the specific implementation of the present application, but the protection scope of the present application is not limited to this. Any changes or replacements within the technical scope disclosed by the present application can be easily conceived by the person skilled in the art, and should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A shadow detection method based on a BSD perspective, characterized in that, include: During the vehicle startup phase, acquire BSD images captured by the onboard BSD camera within the current time period; The first segmentation model is used to perform semantic segmentation on the shadow region of the BSD image to obtain the shadow region segmentation results of multiple frames in the current time period; After the vehicle enters normal driving mode, the real-time acquired BSD image and the multi-frame shadow region segmentation results are input into the second segmentation model, and the change in the segmentation result of the current frame relative to the previous frame is predicted by the second segmentation model. The shadow region segmentation result of the previous frame obtained from the multi-frame shadow region segmentation results is superimposed with the change amount of the segmentation result to obtain the shadow region detection result of the current frame. The first segmentation model uses a large model, while the second segmentation model uses a small model.

2. The shadow detection method based on BSD perspective according to claim 1, characterized in that, The first segmentation model is used to perform semantic segmentation on the shadow region of the preprocessed BSD image to obtain the shadow region segmentation results of multiple frames in the current time period, including: The first segmentation model is used to perform semantic segmentation on the shadow region of the preprocessed BSD image to obtain the shadow region segmentation results of at least the first 10 frames in the current time period; Save the shadow region segmentation results of at least the first 10 frames to the continuous frame segmentation result pool.

3. The shadow detection method based on BSD perspective according to claim 2, characterized in that, Obtaining the shadow region detection results for the current frame also includes: The shadow region detection result of the current frame is saved to the continuous frame segmentation result pool and input into the second segmentation model as the shadow region segmentation result of the next frame.

4. The shadow detection method based on BSD perspective according to claim 2, characterized in that, The first segmentation model is used to perform semantic segmentation on the shadow region of the preprocessed BSD image to obtain the shadow region segmentation results of multiple frames in the current time period, and the method also includes: The shadow region segmentation result is represented as a matrix with elements 0 and 1. Element 0 indicates that the region in the BSD image is not a shadow, and element 1 indicates that the region in the BSD image is a shadow.

5. The shadow detection method based on BSD perspective according to claim 4, characterized in that, The change in segmentation result of the current frame relative to the previous frame is obtained by predicting using the second segmentation model, including: The change in the segmentation result is represented as a matrix of elements True and False. The element True indicates that the element of the current frame in the BSD image is the same as the element of the previous frame in the same region, and the element False indicates that the element of the current frame in the BSD image is not the same as the element of the previous frame in the same region.

6. The shadow detection method based on BSD perspective according to claim 5, characterized in that, The shadow region segmentation result of the previous frame, obtained from the multi-frame shadow region segmentation results, is superimposed with the change in the segmentation result to obtain the shadow region detection result of the current frame, including: Based on the element values ​​of the matrix of changes in the segmentation results, the shadow region segmentation results of the previous frame are modified; If the element value of the matrix of changes in the segmentation result is True, then the shadow region segmentation result of the previous frame is the shadow region detection result of the current frame. If the element value of the matrix of changes in the segmentation result is False, then the shadow region segmentation result of the previous frame is flipped to obtain the shadow region detection result of the current frame.

7. The shadow detection method based on BSD perspective according to claim 1, characterized in that, During the vehicle startup phase, acquiring BSD images captured by the onboard BSD camera within the current time period also includes: The BSD image is preprocessed; The preprocessing process includes: image scaling, image color mapping, and image normalization.

8. A shadow detection system based on a BSD perspective, characterized in that, include: The image acquisition module is used to acquire BSD images captured by the vehicle's BSD camera during the current time period during the vehicle's startup phase. The first model processing module is used to perform semantic segmentation on the shadow region of the BSD image using a first segmentation model to obtain the shadow region segmentation results of multiple frames in the current time period. The second model processing module is used to input the real-time acquired BSD image and the multi-frame shadow region segmentation results into the second segmentation model after the vehicle enters the normal driving state, and to predict the change in the segmentation result of the current frame relative to the previous frame through the second segmentation model. The detection result output module is used to superimpose the shadow region segmentation result of the previous frame obtained from the multi-frame shadow region segmentation result with the change amount of the segmentation result to obtain the shadow region detection result of the current frame. The first segmentation model uses a large model, while the second segmentation model uses a small model.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the shadow detection method based on the BSD perspective as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that performs the shadow detection method based on the BSD perspective as described in any one of claims 1 to 7.

Images