Human body detection method, system and device based on RGB-D image and medium
By combining human detection methods based on RGB-D images with foreground segmentation and cluster analysis of Depth images, the problem of inaccurate detection using RGB images is solved, achieving efficient human detection in occluded scenes.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHENZHEN ORBBEC CO LTD
- Filing Date
- 2022-11-25
- Publication Date
- 2026-07-10
AI Technical Summary
Existing human detection methods based on RGB images are prone to false positives or false negatives when people or objects occlude each other, resulting in inaccurate detection.
Human detection is performed using RGB-D images. Foreground segmentation is performed on the depth image, clustering results are obtained using clustering analysis rules, and human targets are detected by combining RGB images with a pre-set RGB human detection network.
It improves the accuracy of pedestrian detection in occluded scenes, reduces false positives and false negatives, and enhances the accuracy and efficiency of detection.
Smart Images
Figure CN115862061B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of multimedia technology, specifically to a method, system, device, and medium for human body detection based on RGB-D images. Background Technology
[0002] Pedestrian detection technology is widely used in information statistics and related emergency measures in public places, such as pedestrian flow control, shopping mall layout reference, and public security. This technology not only has broad application prospects and great potential in intelligent monitoring systems, but it is also an attractive and challenging problem in computer vision. Visual analysis of pedestrian movement is an emerging and cutting-edge research field, involving multiple areas such as intelligent assisted driving, motion capture, intelligent monitoring, human behavior recognition and analysis, and environmental control and monitoring.
[0003] In public places with high pedestrian traffic, the mutual obstruction between people and between people and objects can easily lead to false detections or missed detections during the pedestrian detection process, resulting in inaccurate detection.
[0004] Therefore, existing technologies need to be improved. Summary of the Invention
[0005] The main objective of this invention is to propose a human body detection method, system, device, and medium based on RGB-D images, so as to at least solve the technical problem of inaccuracy in existing human body detection methods.
[0006] A first aspect of the present invention provides a human detection method based on RGB-D images, comprising: acquiring an RGB-D image containing a human target; wherein the RGB-D image includes an RGB image and a depth image; performing foreground segmentation on the depth image to obtain a foreground image; processing the foreground image according to a preset clustering analysis rule to obtain n clustering results; wherein n is an integer greater than or equal to 0; processing the RGB image according to the n clustering results to obtain n RGB sub-images; and performing human detection on the RGB sub-images using a preset RGB human detection network to obtain a detection result of the human target.
[0007] A second aspect of the present invention provides a human detection device based on RGB-D images, comprising: an image acquisition unit for acquiring an RGB-D image containing a human target; wherein the RGB-D image includes an RGB image and a depth image; a foreground segmentation unit for performing foreground segmentation on the depth image to obtain a foreground image; a clustering analysis unit for obtaining n clustering results based on the foreground image and preset clustering analysis rules; wherein n is an integer greater than or equal to 0; an image processing unit for processing the RGB image based on the n clustering results to obtain n RGB sub-images; and a target detection unit for performing human detection on the RGB sub-images through a preset RGB human detection network to obtain detection results of the human target.
[0008] A third aspect of the present invention provides a human body detection system, the human body detection system including a depth camera and a processor; the depth camera is used to acquire RGB-D images related to a target; wherein the RGB-D images include RGB images and depth images; the processor is used to perform the steps of the human body detection method of the first aspect.
[0009] In a fourth aspect, the present invention provides an electronic device including a memory, a processor, and a bus; the bus is used to enable communication between the memory and the processor; the processor is used to execute a computer program stored in the memory; when the processor executes the computer program, it implements the steps in the human body detection method provided in the first aspect.
[0010] A fifth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that, when the computer program is executed by a processor, it implements the steps of the human body detection method provided in the first aspect.
[0011] This invention provides a human detection method, system, device, and medium based on RGB-D images. It acquires RGB-D images of the target using a depth camera, performs foreground segmentation on the depth image to obtain a foreground image, and obtains n clustering results based on the foreground image and preset clustering analysis rules. Based on the clustering results, the RGB image, and a preset RGB human detection network, the target detection result is obtained. When applied to pedestrian detection in public places, this method considers not only the RGB image but also the depth image within the RGB-D image as a reference. By combining the advantages of both RGB and depth images, false detections caused by light and shadows in pure RGB images can be avoided. Furthermore, by utilizing the clustering results, RGB image, and preset RGB human detection network, the segmented human body can be fed into the RGB human detection network, reducing interference from other pedestrians and background. Therefore, the detection method using RGB-D images, i.e., a multimodal approach, can effectively improve the accuracy of pedestrian detection in occluded scenes and reduce false positives and false negatives. Attached Figure Description
[0012] To more clearly illustrate the technical solutions in the embodiments of this application 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 recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0013] Figure 1 This is a schematic diagram of the human body detection system provided in an embodiment of the present invention;
[0014] Figure 2 This is a flowchart illustrating the human detection method based on RGB-D images provided in an embodiment of the present invention.
[0015] Figure 3 This is a schematic diagram of the process for obtaining clustering results in the human detection method based on RGB-D images provided in an embodiment of the present invention;
[0016] Figure 4 This is a schematic diagram of projecting a foreground image onto a ground plane and obtaining scattered points projected onto the ground plane in an embodiment of the present invention;
[0017] Figure 5 This is a schematic diagram showing that different categories of scatter points correspond to different colors in an embodiment of the present invention;
[0018] Figure 6 This is a schematic diagram showing the different clustering results in the embodiments of the present invention;
[0019] Figure 7This is a schematic diagram of the structure of the human body detection device based on RGB-D images provided in an embodiment of the present invention;
[0020] Figure 8 This is a schematic diagram of the module connections inside an electronic device provided in an embodiment of the present invention.
[0021] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation
[0022] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
[0023] It should be noted that related terms such as "first" and "second" can be used to describe various components, but these terms do not limit the component. These terms are only used to distinguish one component from another. For example, without departing from the scope of the invention, the first component can be referred to as the second component, and the second component can similarly be referred to as the first component. The term "and / or" refers to any one or more combinations of related and descriptive terms.
[0024] Existing human detection methods, especially those based on RGB, almost always analyze RGB images without considering multimodal information, such as depth information. This means that detection methods using only RGB images still suffer from false positives and false negatives, resulting in low detection accuracy.
[0025] Some human detection methods that consider depth information often do not analyze all pixel information in the image due to the large amount of image computation. This results in low detection accuracy for detection methods that miss some pixel information.
[0026] Please see Figure 1 , Figure 1 The human detection system 10 provided in this embodiment of the invention includes a depth camera 11 and a processor 12. The depth camera 11 is used to acquire RGB-D images containing human targets. The RGB-D images include RGB images and depth images. The processor 12 is used to execute a human detection method based on RGB-D images to obtain the detection results of human targets.
[0027] The depth camera 11 includes structured light depth cameras, binocular depth cameras, time-of-flight depth cameras, and depth cameras with RGB cameras (i.e., RGB-D cameras), which can be used to acquire depth images (i.e., depth images in RGB-D images). In practical applications, when the depth camera is installed in a shopping mall area, the human detection method of this embodiment can be used to detect the flow of people in the shopping mall area.
[0028] Please see Figure 2 , Figure 2 This is a schematic flowchart of a human detection method based on RGB-D images provided in an embodiment of the present invention. The human detection method provided in this embodiment includes the following steps:
[0029] Step S210: Obtain an RGB-D image containing the human target.
[0030] An RGB-D image can include both an RGB image and a Depth image. Essentially, an RGB-D image consists of two images: a standard RGB three-channel color image and a Depth image. The RGB image describes the appearance, color, and texture information of an object, while the Depth image describes the spatial depth information of the object.
[0031] The following explains the definitions and relationships between RGB images and Depth images:
[0032] The RGB color mode of RGB images is an industry color standard. It obtains a variety of colors by changing the three color channels of red (R), green (G), and blue (B) and superimposing them. RGB represents the colors of the three channels of red, green, and blue. This standard includes almost all colors that human vision can perceive and is one of the most widely used color systems.
[0033] In 3D computer graphics, a depth image is an image or image channel that contains information about the distances to the surfaces of scene objects from the viewpoint. A depth image is similar to a grayscale image, except that each pixel value represents the actual distance from the sensor to the object. Typically, RGB images and depth images are registered, resulting in a one-to-one correspondence between pixels.
[0034] It should be understood that when acquiring RGB-D images of a target using a depth camera, the acquired RGB-D images can be RGB-D images corresponding to multiple consecutive frames. For example, if the acquired image is a 20-second long sequence of 200 RGB-D images covering 200 frames (where each frame is a still image), then the 200 RGB-D images are RGB-D images corresponding to consecutive frames. Since RGB-D images can include both RGB and depth images, what is actually acquired are RGB and depth images corresponding to multiple consecutive frames. When the first RGB image is obtained, the corresponding first depth image is also obtained simultaneously. Therefore, in subsequent steps S220, S230, S240, and S250, the object being analyzed can be just one frame of the RGB-D image.
[0035] Step S220: Perform foreground segmentation on the Depth image to obtain the foreground image.
[0036] Specifically, depth images contain background and foreground components. When foreground segmentation is performed on a depth image, the background and foreground can be separated to obtain the foreground image. The obtained foreground image often has a stronger correlation with the target, which helps reduce the amount of information processing during detection. At the same time, foreground segmentation can extract the background image, which means that the interference of the background image can be avoided in subsequent processing.
[0037] Step S230: Process the foreground image using preset clustering analysis rules to obtain n clustering results;
[0038] In this embodiment, after obtaining the foreground image, the correlation between the foreground image and the target is stronger, thus the effectiveness of human detection on the foreground image is higher. By processing the foreground image through preset clustering analysis rules, n clustering results can be obtained; each of the n clustering results corresponds to a target object. Since the target objects appearing in the image may be human bodies, they may also include non-human targets such as boxes, shopping carts, etc. due to the complexity of the environment; that is, the clustering result represents all target objects detected within a preset range.
[0039] It should be understood that clustering can process the foreground image to obtain n clustering results. Therefore, when the depth camera in this embodiment is applied to the shopping mall area, the n clustering results obtained are the results obtained after analyzing all pixels in the foreground image. That is, all clustering results can reflect all targets that can be covered in the foreground image. Although it has a certain amount of computation, there will be no problem of missed detection.
[0040] Step S240: Process the RGB image based on the n clustering results to obtain n RGB sub-images.
[0041] In this embodiment, n clustering results are mapped to RGB images, or the RGB images are segmented based on the n clustering results, with each clustering result corresponding to an RGB sub-image, resulting in n RGB sub-images.
[0042] Step S250: Perform human detection on the RGB sub-image using a preset RGB human detection network to obtain the detection result of the human target.
[0043] In this embodiment, the RGB sub-image is input into a preset RGB human detection network. Human detection is performed on the target in the RGB sub-image to determine whether the image in the RGB sub-image is a human target, thereby eliminating interference from non-human targets. Based on the clustering results, the image is then mapped back to the RGB image, or the clustered targets are segmented, and the resulting RGB sub-image is sent to the RGB human detection network. This can achieve the technical effect of reducing interference from other pedestrians, backgrounds, etc.
[0044] After obtaining n RGB sub-images, the target detection results corresponding to the n RGB sub-images can be quickly and accurately analyzed based on a pre-trained RGB human detection network. Here, the RGB human detection network represents a neural network model, such as a trained neural network model, which can use computer vision technology to determine whether a target in an image is a human body.
[0045] Through the implementation of the above embodiments, firstly, by adding a depth image to the RGB image, problems that pure RGB images cannot solve can be addressed to a certain extent. For example, using a depth image can avoid the influence of lighting and shadows. Secondly, by performing cluster analysis on the foreground image obtained after foreground segmentation of the depth image, without including the background image, the amount of information processing during detection can be reduced. Finally, based on the trained RGB human detection network, no actual calculation is required; only automated detection and analysis of the model are involved, which can quickly and accurately obtain the detection results of the target, thereby improving the accuracy of pedestrian detection and reducing false positives and false negatives.
[0046] In this embodiment, the step of performing foreground segmentation on the Depth image to obtain a foreground image specifically includes: performing foreground segmentation based on the pixel values in the Depth image to obtain a foreground image.
[0047] Specifically, the separation process includes morphological operations and background modeling. Morphological operations are used to change the shape of targets in the depth image, while background modeling involves building a background model that can be used to detect targets in the depth image. In other words, morphological operations and background modeling can separate the background and foreground images of the depth image to obtain the foreground image. Generally, the foreground image contains the target, while the background image does not. Therefore, separating the foreground image from the depth image effectively reduces the amount of image information that needs to be processed.
[0048] Please see Figure 3 and Figure 4 , Figure 3 This diagram illustrates the process of obtaining clustering results in the human detection method based on RGB-D images provided in this embodiment of the invention. Figure 4 This diagram illustrates the projection of a foreground image onto a ground plane to obtain corresponding scattered points. In this embodiment, the steps for obtaining n clustering results based on the foreground image and preset clustering analysis rules specifically include:
[0049] Step S231: Perform plane fitting on the Depth image using a preset fitting algorithm to obtain the fitted ground plane.
[0050] Specifically, methods for fitting the plane to the Depth image include using the RANSAC algorithm and the least squares algorithm. The fitted ground plane can be obtained through these algorithms.
[0051] Step S232: Project the foreground image onto the ground plane to obtain scattered points projected onto the ground plane.
[0052] Specifically, since foreground images often contain various targets, such as people, pets, and obstacles, when the foreground image is projected onto the ground plane, the resulting scatter points are the points that make up the various targets mentioned above. The beneficial effects of projecting the foreground image onto the ground plane and obtaining the scatter points projected onto the ground plane are: on the one hand, it can reduce the influence of pedestrian posture, and on the other hand, it can use these scatter points to distinguish different targets.
[0053] Step S233: Perform cluster analysis on the scatter points to obtain n cluster results.
[0054] For details, please refer to Figure 5Different categories of scatter points can be distinguished using different colors, and different colored scatter points are used to match corresponding targets. After cluster analysis is performed on the scatter points to obtain the distinguished targets, each target corresponds to a scatter point of a different color. In other words, cluster analysis primarily refers to the categories of scatter points to distinguish the corresponding target objects. After obtaining the distinguished target objects, n clustering results can be obtained based on the target objects and the depth image. For example, the target objects can be mapped onto the depth image. Different clustering results can be found in [reference needed]. Figure 6 , Figure 6 It includes two clustering results (one clustering result is...) Figure 6 The white area in the image, another clustering result is... Figure 6 The black area in the middle represents the region in the Depth image, which is the clustering result.
[0055] In this embodiment, the step of obtaining n clustering results based on the target object and the depth image specifically includes: mapping different target objects to the depth image to obtain segmented depth maps, and obtaining n clustering results based on the depth maps.
[0056] Specifically, after different target objects are obtained, i.e., after differentiation is completed, different target objects are mapped back to the Depth image according to the mapping relationship, resulting in a segmented depth map. The segmented depth map includes n clustering results, meaning that n clustering results can be obtained from the depth map. It's important to note that the "segmented depth map" is not directly equivalent to the Depth image; rather, it should be understood as the image generated after different target objects are mapped back to the Depth image.
[0057] In this embodiment, after the step of inputting the RGB sub-image into a preset RGB human detection network to perform human detection and obtain the detection result of the target, the method further includes: if the detection result is that the target belongs to a non-human body, then change the label corresponding to the target; if the detection result is that the target belongs to a human body, then maintain the label corresponding to the target.
[0058] Specifically, upon receiving the detection result, if the result indicates that the target is not a human being (i.e., the target is not a person, but could be a shadow, cargo, obstacle, suitcase, backpack, or other non-human object), the marker corresponding to the target is changed, and the non-human target is no longer detected or tracked. If the result indicates that the target is a human being, the marker corresponding to the target is maintained. The marker can be a pre-set graphic on the target, designed to facilitate subsequent tracking of each target. In one specific embodiment, the human detection system is used to monitor a certain area and record whether anyone passes through that area. Since monitoring is a long-term process, when data analysis of the video data collected by the human detection system is needed, especially when extracting images where a human body appears in the video data, using an RGB-D image-based human detection method to detect human bodies in the image frames of the video data and marking the image frames containing human bodies in the detection results facilitates the rapid extraction of images where a human body appears in the video data.
[0059] In this embodiment, after acquiring the RGB-D image of the target captured by the depth camera, the method further includes: analyzing the RGB-D image to obtain the number of targets in the RGB-D image; comparing the number of targets with a preset target number threshold; and when the comparison result shows that the number of targets is greater than the preset target number threshold, performing foreground segmentation on the depth image to obtain a foreground image.
[0060] After obtaining the RGB-D image, the RGB-D image is analyzed to obtain the number of targets in the RGB-D image. The number of targets is compared with a preset target number threshold. When the comparison result is that the number of targets is greater than the preset target number threshold, it indicates that there are many targets in the RGB-D image. When performing human detection, it is easy to miss or be inaccurate due to overlap and shadows. Therefore, this part of the RGB-D image is identified, and subsequent steps S220, S230, S240 and S250 are performed specifically on this part of the RGB-D image.
[0061] Meanwhile, when the comparison result shows that the number of targets is less than or equal to the preset target number threshold, it indicates that there are fewer targets in the RGB-D image and there is generally no human body overlap problem. In other words, there is no need to analyze through the Depth image, and the detection result can be obtained by analyzing the regular RGB image. This helps to reduce the detection workload of the human body detection method in this embodiment, that is, to reduce the processing load of the human body detection system.
[0062] In addition, the present invention also provides a human body detection system, which includes a depth camera and a processor; wherein the depth camera is used to acquire RGB-D images related to the target, and the processor is used to execute the steps in the human body detection method provided in the above embodiments.
[0063] Therefore, when applied to public places for pedestrian detection, this invention considers not only RGB images but also the depth image from the RGB-D image as a reference. By combining the advantages of both RGB and depth images, it can solve problems that pure RGB images cannot address to some extent. For example, using depth information is unaffected by lighting or shadows. Secondly, clustering using the depth image first segments targets with different spatial relationships, reducing the number of targets requiring analysis. Finally, based on the clustering results, the RGB image, and a pre-trained RGB human detection network, the detection results are obtained. Since the RGB human detection network is pre-trained, detection efficiency and accuracy are significantly improved. Furthermore, mapping the clustering results back to the RGB image effectively segments occluded human bodies, and feeding these segmented bodies into the RGB human detection network reduces interference from other pedestrians and background elements. Furthermore, when the target is identified as a human body, this invention accurately calculates the target's actual center of gravity using depth images and obtains a more accurate predicted center of gravity. Based on the actual and predicted center of gravity, an accurate and reliable target tracking strategy can be formulated, ultimately improving the stability and accuracy of target tracking. In other words, when applied to public places such as shopping malls and supermarkets, this invention not only has high accuracy in detecting crowds but is also suitable for precise tracking and positioning of people.
[0064] Figure 7 This is a schematic diagram of a human body detection device for RGB-D images provided in an embodiment of this application. The included units are used to perform... Figure 2 The steps in the corresponding embodiments. Please refer to the details. Figure 2 The relevant descriptions in the corresponding embodiments are shown below. For ease of explanation, only the parts relevant to this embodiment are shown. See also... Figure 7 The human body detection device 70 includes:
[0065] Image acquisition unit 71 is used to acquire an RGB-D image containing a human target; wherein the RGB-D image includes an RGB image and a depth image;
[0066] Foreground segmentation unit 72 is used to segment the Depth image for the foreground to obtain a foreground image;
[0067] Clustering analysis unit 73 is used to obtain n clustering results based on the foreground image and preset clustering analysis rules; where n is an integer greater than or equal to 0;
[0068] The image processing unit 74 is used to process the RGB image based on n clustering results to obtain n RGB sub-images;
[0069] The target detection unit 75 is used to perform human detection on the RGB sub-image through a preset RGB human detection network to obtain the detection result of the human target.
[0070] Figure 8 An electronic device provided in an embodiment of the present invention is shown. This electronic device can be used to implement the human detection method based on RGB-D images in any of the foregoing embodiments. The electronic device includes:
[0071] The system includes a memory 81, a processor 82, a bus 83, and a computer program stored on the memory 81 and executable on the processor 82. The memory 81 and the processor 82 are connected via the bus 83. When the processor 82 executes the computer program, it implements the human body detection method based on RGB-D images described in the foregoing embodiments. The number of processors can be one or more.
[0072] The memory 81 can be a high-speed random access memory (RAM) or a non-volatile memory, such as a disk storage device. The memory 81 is used to store executable program code, and the processor 82 is coupled to the memory 81.
[0073] Furthermore, embodiments of this application also provide a computer-readable storage medium, which may be disposed in the electronic device in the above embodiments, and the computer-readable storage medium may be a memory.
[0074] The computer-readable storage medium stores a computer program that, when executed by a processor, implements the human body detection method based on RGB-D images described in the foregoing embodiments. Furthermore, the computer-readable storage medium can also be a USB flash drive, external hard drive, read-only memory (ROM), RAM, magnetic disk, or optical disk, or any other medium capable of storing program code.
[0075] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.
[0076] The modules described as separate components may or may not be physically separate. Similarly, the components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment, depending on actual needs.
[0077] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated modules described above can be implemented in hardware or as software functional modules.
[0078] If the integrated module is implemented as a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a readable storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned readable storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.
[0079] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.
[0080] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.
[0081] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.
Claims
1. A human detection method based on RGB-D images, characterized in that, include: Acquire an RGB-D image containing a human target; wherein the RGB-D image includes an RGB image and a depth image; Perform foreground segmentation on the Depth image to obtain a foreground image; The Depth image is fitted to a plane using a preset fitting algorithm to obtain a fitted ground plane. The foreground image is then projected onto the ground plane to obtain scattered points projected onto the ground plane. Cluster analysis is performed on the scattered points to obtain n clustering results; where n is an integer greater than or equal to 0; The RGB image is processed based on the n clustering results to obtain n RGB sub-images; Human detection is performed on the RGB sub-image using a preset RGB human detection network to obtain the detection result of the human target.
2. The human detection method based on RGB-D images as described in claim 1, characterized in that, The step of performing foreground segmentation on the Depth image to obtain a foreground image includes: Foreground segmentation is performed based on the pixel values in the Depth image to obtain the foreground image.
3. The human detection method based on RGB-D images as described in claim 1, characterized in that, The cluster analysis of the scatter points yields n clustering results, including: The different scattered points are mapped to the Depth image to obtain the segmented depth map; Based on the depth map, n clustering results are obtained.
4. The human detection method based on RGB-D images as described in claim 1, characterized in that, After the step of performing human detection on the RGB sub-image using a preset RGB human detection network to obtain the detection result of the human target, the method further includes: If the detection result indicates that the target is not a human, then the label corresponding to the target is changed; If the detection result indicates that the target belongs to the human body, then the label corresponding to the target is maintained.
5. The human detection method based on RGB-D images as described in claim 1, characterized in that, Following the step of acquiring the RGB-D image containing the human target, the method further includes: The number of targets in the RGB-D image is obtained by analyzing the RGB-D image. The target number is compared with a preset target number threshold. When the comparison result shows that the target number is greater than the preset target number threshold, the step of performing foreground segmentation on the Depth image to obtain a foreground image is performed.
6. A human body detection device based on RGB-D images, characterized in that, include: An image acquisition unit is used to acquire an RGB-D image containing a human target; wherein the RGB-D image includes an RGB image and a depth image; A foreground segmentation unit is used to perform foreground segmentation on the Depth image to obtain a foreground image; The clustering analysis unit is used to perform planar fitting on the Depth image using a preset fitting algorithm to obtain a fitted ground plane, project the foreground image onto the ground plane to obtain scattered points projected onto the ground plane, and perform clustering analysis on the scattered points to obtain n clustering results; where n is an integer greater than or equal to 0; An image processing unit is configured to process the RGB image based on the n clustering results to obtain n RGB sub-images; The target detection unit is used to perform human detection on the RGB sub-image through a preset RGB human detection network to obtain the detection result of the human target.
7. A human body detection system, characterized in that, Including depth cameras and processors; The depth camera is used to acquire RGB-D images related to the target; wherein, the RGB-D images include RGB images and depth images; The processor is used to perform the steps in the human body detection method as described in any one of claims 1 to 5.
8. An electronic device, characterized in that, Includes memory, processor, and bus; The bus is used to enable communication between the memory and the processor; The processor is used to execute computer programs stored in the memory; When the processor executes the computer program, it implements the steps in the human body detection method according to any one of claims 1 to 5.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the human body detection method according to any one of claims 1 to 5.