Human back bone key point detection model construction method and key point detection method

Abstract

The application relates to a human back bone key point detection model construction method and device, computer equipment and a storage medium, wherein the method comprises the following steps: preprocessing sample human back three-dimensional topography data to obtain a sample normal rendering graph, obtaining the positions of labeled bone key points in the sample normal rendering graph, training and adjusting an initial neural network model based on the sample normal rendering graph and the labeled bone key point positions, and obtaining a human back bone key point detection model through training, so that the obtained model can support subsequent accurate detection of human back bone key points. In addition, the application also provides a human back bone key point detection method and device based on the human back bone key point detection model constructed above, computer equipment and a storage medium, which can accurately detect human back bone key points.

Description

Technical Field

[0001] This application relates to the field of image processing technology based on artificial intelligence, and in particular to a method, apparatus, computer equipment and storage medium for constructing a detection model of bony key points on the human back, as well as a method, apparatus, computer equipment and storage medium for detecting bony key points on the human back. Background Technology

[0002] Locating key bony points on the human back is crucial for analyzing back symmetry and quantitatively assessing the impact of scoliosis on posture using color images of the back.

[0003] Currently, the mainstream image key point detection methods are based on image feature information (such as SIFT and other feature vector detection methods), which have good processing results for objects with rich image feature information such as grayscale, texture, and shape.

[0004] If the above-mentioned key point detection method based on image feature information is directly applied to detect bony key points on the human back, the bony key point features are not obvious in color photographs, which will seriously affect the accuracy of bony key point localization and result in low accuracy of bony key point detection on the human back. Summary of the Invention

[0005] Therefore, it is necessary to address the aforementioned technical problems by providing a method, apparatus, computer equipment, and storage medium for constructing a human back bony key point detection model that supports high-precision detection, as well as a method, apparatus, computer equipment, and storage medium for detecting human back bony key points that can achieve high-precision detection.

[0006] A method for constructing a detection model of bony key points in the human back, the method comprising:

[0007] Obtain three-dimensional topographic data of the human back sample;

[0008] Preprocess the three-dimensional topographic data of the human back sample to obtain the sample normal rendering map;

[0009] Obtain the locations of the marked bony key points in the sample normal rendering image;

[0010] Based on the sample normal rendering map and the labeled bony key point locations, and with accuracy as the optimization objective, an initial neural network model is trained to obtain a human back bony key point detection model.

[0011] In one embodiment, obtaining the three-dimensional morphological data of the back of a human sample includes:

[0012] By using a 3D depth sensing component, the imaging area on the back of multiple sample objects is scanned to obtain 3D morphological data of the human back of the sample.

[0013] In one embodiment, preprocessing the three-dimensional topographic data of the human back sample to obtain the sample normal rendering map includes:

[0014] Principal component analysis was used to correct the orientation of the three-dimensional topographic data of the human back sample, and a normal rendering map of the sample was generated.

[0015] In one embodiment, principal component analysis is used to perform orientation correction on the three-dimensional topographic data of the human back sample to generate a sample normal rendering map, including:

[0016] Principal component analysis was used to perform orientation correction on the three-dimensional topography data of the human back sample to generate an initial sample normal rendering map.

[0017] The initial sample normal rendering map is then subjected to back region cropping and data augmentation to obtain the sample normal rendering map.

[0018] In one embodiment, obtaining the locations of the marked bony key points in the sample normal rendering map includes:

[0019] Push the sample normal rendering map to a third-party expert annotation platform;

[0020] Receive the sample normal rendering image of the labeled bony key points from the third-party expert annotation platform, and obtain the location of the labeled bony key points in the sample normal rendering image.

[0021] A device for constructing a detection model of bony key points in the human back, the device comprising:

[0022] The sample data acquisition module is used to acquire three-dimensional morphological data of the back of a human sample.

[0023] The preprocessing module is used to preprocess the three-dimensional topographic data of the human back sample to obtain the sample normal rendering map;

[0024] The annotation acquisition module is used to obtain the locations of the annotated bony key points in the sample normal rendering map;

[0025] The model training module is used to train an initial neural network model based on the sample normal rendering map and the labeled bony key point locations, with accuracy as the optimization objective, to obtain a human back bony key point detection model.

[0026] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:

[0027] Obtain three-dimensional topographic data of the human back sample;

[0028] Preprocess the three-dimensional topographic data of the human back sample to obtain the sample normal rendering map;

[0029] Obtain the locations of the marked bony key points in the sample normal rendering image;

[0030] Based on the sample normal rendering map and the labeled bony key point locations, and with accuracy as the optimization objective, an initial neural network model is trained to obtain a human back bony key point detection model.

[0031] A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, performs the following steps:

[0032] Obtain three-dimensional topographic data of the human back sample;

[0033] Preprocess the three-dimensional topographic data of the human back sample to obtain the sample normal rendering map;

[0034] Obtain the locations of the marked bony key points in the sample normal rendering image;

[0035] Based on the sample normal rendering map and the labeled bony key point locations, and with accuracy as the optimization objective, an initial neural network model is trained to obtain a human back bony key point detection model.

[0036] The aforementioned method, apparatus, computer equipment, and storage medium for constructing a human back bony keypoint detection model preprocesses the three-dimensional topographic data of a sample human back to obtain a sample normal rendering map. The locations of the bony keypoints already marked in the sample normal rendering map are then obtained. Based on the sample normal rendering map and the marked bony keypoint locations, an initial neural network model is trained and adjusted, resulting in a human back bony keypoint detection model. Throughout this process, on the one hand, obtaining the marked bony keypoint locations in the sample normal rendering map makes the image features of the human back bony keypoints more prominent; on the other hand, by employing a neural network model with accuracy as the optimization objective, the parameters in the neural network model are adjusted. Therefore, the resulting model can support the subsequent accurate detection of human back bony keypoints.

[0037] In addition, this application also provides a method for detecting bony key points of the human back, the method comprising:

[0038] Acquire three-dimensional topographic data of the human back;

[0039] Normal rendering is performed on the three-dimensional topographic data of the human back to obtain a normal rendering map.

[0040] The normal rendering map is input into the human back bony key point detection model to obtain the bony key point positions in the normal rendering map. The human back bony key point detection model is constructed by the above method.

[0041] Based on the location of bony key points in the normal rendering image, the location of bony key points on the human back can be obtained.

[0042] In one embodiment, the location of bony key points on the human back is obtained based on the location of bony key points in the normal rendering image, including:

[0043] Obtain the coordinate correspondence between pixel coordinates in the normal rendering image and the coordinates of the 3D topography data of the human back;

[0044] Based on the location and coordinate correspondence of bony key points in the normal rendering image, the three-dimensional coordinates of bony key points on the human back are obtained.

[0045] In addition, this application also provides a device for detecting bony key points of the human back, the device comprising:

[0046] The 3D data acquisition module is used to acquire 3D topographic data of the human back.

[0047] The normal rendering module is used to perform normal rendering processing on the three-dimensional shape data of the human back to obtain a normal rendering map.

[0048] The model detection module is used to input the normal rendering map into the human back bony key point detection model to obtain the bony key point positions in the normal rendering map. The human back bony key point detection model is constructed by the above method.

[0049] The location acquisition module is used to obtain the location of the bony key points on the human back based on the location of the bony key points in the normal rendering map.

[0050] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:

[0051] Acquire three-dimensional topographic data of the human back;

[0052] Normal rendering is performed on the three-dimensional topographic data of the human back to obtain a normal rendering map.

[0053] The normal rendering map is input into the human back bony key point detection model to obtain the bony key point positions in the normal rendering map. The human back bony key point detection model is constructed by the above method.

[0054] Based on the location of bony key points in the normal rendering image, the location of bony key points on the human back can be obtained.

[0055] A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, performs the following steps:

[0056] Acquire three-dimensional topographic data of the human back;

[0057] Normal rendering is performed on the three-dimensional topographic data of the human back to obtain a normal rendering map.

[0058] The normal rendering map is input into the human back bony key point detection model to obtain the bony key point positions in the normal rendering map. The human back bony key point detection model is constructed by the above method.

[0059] Based on the location of bony key points in the normal rendering image, the location of bony key points on the human back can be obtained.

[0060] The aforementioned method, apparatus, computer equipment, and storage medium for detecting bony key points on the human back involve normal rendering processing of the three-dimensional topographic data of the human back to obtain a normal rendering image. This normal rendering image is then input into the human back bony key point detection model, and the positions of the bony key points on the human back are obtained based on the locations of the bony key points in the normal rendering image. Throughout this process, on the one hand, normal rendering processing of the three-dimensional topographic data of the human back makes the image features of the bony key points on the human back more prominent; on the other hand, the human back bony key point detection model is constructed using the aforementioned method, and this model supports accurate detection of bony key points on the human back. Therefore, the entire solution can achieve accurate detection of bony key points on the human back. Attached Figure Description

[0061] Figure 1 This is a flowchart illustrating a method for constructing a detection model of bony key points on the human back in one embodiment.

[0062] Figure 2 Point cloud map of the three-dimensional topography of the human back;

[0063] Figure 3 Render the sample normal map corresponding to the back of the human body;

[0064] Figure 4 A schematic diagram of the bony key points corresponding to the human back;

[0065] Figure 5 Another embodiment is a flowchart illustrating the method for constructing a detection model of bony key points on the human back.

[0066] Figure 6 This is a schematic diagram of a device for constructing a human back bony key point detection model in one embodiment.

[0067] Figure 7 This is a flowchart illustrating a method for detecting bony key points on the human back in one embodiment.

[0068] Figure 8 This is a schematic diagram of a human back bony key point detection device in one embodiment;

[0069] Figure 9 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0070] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0071] like Figure 1 As shown, this application provides a method for constructing a detection model of bony key points on the human back, the method including:

[0072] S120: Obtain the three-dimensional morphological data of the back of the sample human body.

[0073] The 3D topographic data of the human back can be obtained by acquiring historical 3D topographic data of the human back, or it can be obtained through real-time acquisition. In real-time acquisition, the 3D topographic data of the human back of multiple sample objects can be collected. Specifically, point cloud data of these sample objects can be collected using a depth sensor, and a corresponding point cloud map is drawn based on the point cloud data collected by the sensor. This point cloud map is the acquired 3D topographic data of the human back. See [link to specific point cloud map] for details. Figure 2 When the depth sensor collects point cloud data, it can collect sensor data on the left and right shoulder widths, from the neck up to the buttocks crease, and aggregate this data to obtain a point cloud map of the human back, thus obtaining the three-dimensional shape data of the human back. By performing data collection (sensing) in the above manner for each sample object, the three-dimensional shape data of the sample's human back can be obtained.

[0074] S140: Preprocess the three-dimensional topographic data of the human back of the sample to obtain the sample normal rendering map.

[0075] The 3D topographic data of the human back samples collected in step S120 are preprocessed. Preprocessing may include orientation correction, normal rendering, and data augmentation to obtain a sample normal rendering map. Orientation correction refers to uniformly correcting the 3D topographic data of the human back samples to the same 3D coordinate system to shield data deviations caused by different coordinate systems during data collection. Orientation correction can be performed through principal component analysis to correct the orientation of the 3D data. Normal rendering makes the key features of the human back bones more prominent. Due to the limited number of samples actually collected, data augmentation is needed to increase the types of data and improve the generalization ability of the model. Data augmentation includes left and right flipping and random cropping. The obtained sample normal rendering map can be found in [reference needed]. Figure 3 .

[0076] S160: Obtain the locations of the marked bony key points in the sample normal rendering map.

[0077] like Figure 4 As shown, the bony key points include the upper and lower edges of the left and right scapulae, the two points on the left and right sides of the posterior superior iliac spine, and seven bony key points of the protuberance. The bony key points already marked in the sample normal rendering can be those already included in the 3D topographic data of the sample human back. This data is extracted here. Specifically, experts can manually mark the bony key points in the 3D topographic data of the sample human back during acquisition. Alternatively, the marked bony key points in the sample normal rendering can be obtained by sending the sample normal rendering to experts, who can then manually mark the bony key points directly on the sample normal rendering. For example, the sample normal rendering can be displayed directly on a touch screen terminal, and experts can manually mark the specific bony key points on the display interface. The touch screen terminal records the positions of the manually marked bony key points, and these bony key points are then integrated into the sample normal rendering based on the recorded positions, thus obtaining the marked bony key point positions in the sample normal rendering.

[0078] S180: Based on the sample normal rendering map and the labeled bony key point locations, and with accuracy as the optimization objective, train the initial neural network model to obtain the human back bony key point detection model.

[0079] The initial neural network model is a pre-cached neural network model, which is a general-purpose neural network model. Alternatively, a neural network model specifically designed for training point cloud data can be selected. Using sample normal rendering maps and labeled bony keypoint locations as input data, the model's parameters are continuously adjusted, such as the specific parameters of the loss function, with accuracy as the optimization objective. Through continuous optimization, a human back bony keypoint detection model is obtained.

[0080] The aforementioned method for constructing a human back bony keypoint detection model involves preprocessing the 3D topographic data of a sample human back to obtain a sample normal rendering map. The locations of the bony keypoints already marked in the sample normal rendering map are then obtained. Based on the sample normal rendering map and the marked bony keypoint locations, an initial neural network model is trained and adjusted, resulting in a human back bony keypoint detection model. Throughout this process, on the one hand, obtaining the marked bony keypoint locations in the sample normal rendering map makes the features of the human back bony keypoints more prominent; on the other hand, by employing a neural network model with accuracy as the optimization objective, the parameters in the neural network model are adjusted. Therefore, the resulting model can support accurate subsequent detection of human back bony keypoints.

[0081] In one embodiment, obtaining three-dimensional topographic data of the back of a human sample includes: scanning the imaging area of ​​the back of multiple sample objects using a three-dimensional depth sensing component to obtain three-dimensional topographic data of the back of the human sample.

[0082] Specifically, the three-dimensional depth sensing component can be a three-dimensional depth sensor. A certain number (more) of three-dimensional depth sensors can collect data on the exposed back of the upper body of the sample object. The range of the collected data includes the left and right shoulder widths, from the neck up to the buttock crease down. That is, the imaging area of ​​the back of each sample object is scanned to obtain the three-dimensional shape data of the sample human back.

[0083] like Figure 5 As shown, in one embodiment, S140 includes:

[0084] S142: Principal component analysis is used to correct the orientation of the three-dimensional shape data of the human back of the sample and generate an initial sample normal rendering map.

[0085] S144: Perform back region clipping and data augmentation on the initial sample normal rendering map in sequence to obtain the sample normal rendering map.

[0086] Principal Component Analysis (PCA) is a multivariate statistical analysis method that uses linear transformations to select a smaller number of significant variables from multiple variables. It is also known as principal component analysis. PCA translates the centroids of models corresponding to the 3D morphology data of different human back samples to a unified, pre-defined reference coordinate system (the same standard coordinate system). This effectively corrects the orientation of the 3D morphology data of the human back samples, thus shielding the accuracy of the data from the impact of collecting data in different spatial coordinate systems and significantly improving the accuracy of subsequent model construction. Based on the 3D topographic data of the human back after orientation correction, an initial sample normal rendering map is generated. Further back region cropping and data augmentation are needed for this initial sample normal rendering map. Since scoliosis information is mainly distributed in the back region of the human body, information from other parts of the body is of no help for classification. Therefore, a specific method is needed to crop out the back region; this process is called back region cropping. Specifically, a trained Faster-RCNN can be used for back detection. Due to the limited number of actual samples, data augmentation is required to increase the types of data and improve the model's generalization ability. This process is called data augmentation, which is achieved through methods such as left-right flipping and random cropping.

[0087] In one embodiment, obtaining the locations of the marked bony key points in the sample normal rendering map includes:

[0088] Push the sample normal rendering image to a third-party expert annotation platform; receive the sample normal rendering image with annotated bony key points from the third-party expert annotation platform, and obtain the location of the annotated bony key points in the sample normal rendering image.

[0089] In this embodiment, the sample normal rendering image is pushed to a third-party expert annotation platform. The third-party expert annotation platform manually annotates the received sample normal rendering image to obtain a sample normal rendering image with annotated bony key points. The sample normal rendering image with annotated bony key points is then fed back, thus obtaining the locations of the annotated bony key points in the sample normal rendering image. Specifically, the third-party expert annotation platform refers to a data platform where professional technicians annotate bony key points.

[0090] In addition, such as Figure 6 As shown, this application also provides a device for constructing a detection model of bony key points on the human back, the device comprising:

[0091] The sample data acquisition module 620 is used to acquire three-dimensional morphological data of the back of the human sample.

[0092] The preprocessing module 640 is used to preprocess the three-dimensional topographic data of the human back sample to obtain the sample normal rendering map.

[0093] The annotation acquisition module 660 is used to acquire the locations of the labeled bony key points in the sample normal rendering map; the model training module 680 is used to train an initial neural network model based on the sample normal rendering map and the locations of the labeled bony key points, with accuracy as the optimization objective, to obtain a human back bony key point detection model.

[0094] The aforementioned device for constructing a human back bony keypoint detection model preprocesses the 3D topographic data of a sample human back to obtain a sample normal rendering map. It then identifies the locations of the bony keypoints already marked in the sample normal rendering map. Based on the sample normal rendering map and the marked bony keypoint locations, an initial neural network model is trained and adjusted, resulting in a human back bony keypoint detection model. Throughout this process, on the one hand, obtaining the marked bony keypoint locations in the sample normal rendering map makes the features of the human back bony keypoints more prominent; on the other hand, by employing a neural network model and optimizing for accuracy, the parameters in the neural network model are adjusted. Therefore, the resulting model can support accurate subsequent detection of human back bony keypoints.

[0095] Specific limitations regarding the device for constructing the human back bony key point detection model can be found in the limitations of the method for constructing the human back bony key point detection model mentioned above, and will not be repeated here. Each module in the aforementioned human back bony key point detection model construction device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0096] In addition, such as Figure 7 As shown, this application also provides a method for detecting bony key points of the human back, the method comprising:

[0097] S720: Acquires three-dimensional topographic data of the human back.

[0098] The 3D topographic data of the human back can be acquired in real time or by receiving data from external devices or transmitted over a network. If real-time acquisition is used, a 3D depth sensor can be used to collect the 3D topographic data of the human back.

[0099] S740: Perform normal rendering processing on the three-dimensional topographic data of the human back to obtain a normal rendering map.

[0100] After obtaining the three-dimensional topographic data of the human back, principal component analysis can be performed to correct the orientation and generate a normal contour map; then normal rendering is performed to obtain a normal rendering map.

[0101] S760: Input the normal rendering map into the human back bony key point detection model to obtain the bony key point positions in the normal rendering map. The human back bony key point detection model is constructed by the above method.

[0102] The human back bony key point detection model is constructed from the aforementioned human back bony key point detection model. This model is a neural network model trained based on sample normal rendering maps and labeled bony key point positions. After inputting the normal rendering map at the input end, the human back bony key point detection model can accurately output the bony key point positions in the line rendering map.

[0103] S780: Based on the location of bony key points in the normal rendering image, obtain the location of bony key points on the human back.

[0104] After obtaining the locations of bony key points in the normal rendering image, the positions of these key points on the human back can be clearly seen. If not necessary, the coordinate correspondence between the pixel coordinates in the normal rendering image and the coordinates of the 3D topographic data of the human back can also be obtained. Based on the locations of the bony key points in the normal rendering image and their coordinate correspondence, the 3D coordinates of the bony key points on the human back can be obtained. This allows for direct acquisition of the 3D coordinates of the bony key points on the human back in a 3D coordinate system, avoiding the shortcomings of traditional image-based key point localization, which only provides pixel coordinates and lacks the 3D information of these key points. This supports further diagnosis and detection of the human back.

[0105] The aforementioned method for detecting bony key points on the human back involves performing normal rendering on the 3D topographic data of the human back to obtain a normal rendering image. This normal rendering image is then input into the human back bony key point detection model, and the locations of the bony key points on the human back are obtained based on the positions of the key points in the normal rendering image. Throughout this process, on the one hand, the normal rendering processing of the 3D topographic data of the human back makes the image features of the bony key points more prominent; on the other hand, the human back bony key point detection model, constructed using the aforementioned method, supports accurate detection of bony key points on the human back. Therefore, the entire solution can achieve accurate detection of bony key points on the human back.

[0106] In addition, such as Figure 8 As shown, this application also provides a device for detecting bony key points of the human back, the device comprising:

[0107] The 3D data acquisition module 820 is used to acquire 3D topographic data of the human back.

[0108] The normal rendering module 840 is used to perform normal rendering processing on the three-dimensional shape data of the human back to obtain a normal rendering map.

[0109] The model detection module 860 is used to input the normal rendering map into the human back bony key point detection model to obtain the bony key point positions in the normal rendering map. The human back bony key point detection model is constructed by the above method.

[0110] The position acquisition module 880 is used to obtain the position of the bony key points on the human back based on the position of the bony key points in the normal rendering map.

[0111] The aforementioned human back bony key point detection device performs normal rendering processing on the three-dimensional topographic data of the human back to obtain a normal rendering map. This normal rendering map is then input into the human back bony key point detection model, and the positions of the bony key points on the human back are obtained based on the locations of the bony key points in the normal rendering map. Throughout this process, on the one hand, the normal rendering processing of the three-dimensional topographic data of the human back makes the image features of the human back bony key points more prominent; on the other hand, the human back bony key point detection model is constructed using the above method, and this model supports accurate detection of human back bony key points. Therefore, the entire solution can achieve accurate detection of human back bony key points.

[0112] In one embodiment, the location acquisition module 880 is further used to acquire the coordinate correspondence between the pixel coordinates in the normal rendering image and the coordinates of the three-dimensional shape data of the human back; and to obtain the three-dimensional coordinates of the bony key points of the human back based on the position of the bony key points in the normal rendering image and the coordinate correspondence.

[0113] It should be understood that, although Figure 1 , Figure 5 as well as Figure 7 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 1 , Figure 5 as well as Figure 7 At least some of the steps in the process may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but may be executed at different times. The execution order of these steps or stages is not necessarily sequential, but may be executed in turn or alternately with other steps or at least some of the steps or stages in other steps.

[0114] Specific limitations regarding the human back bony key point detection device can be found in the limitations of the human back bony key point detection method described above, and will not be repeated here. Each module in the aforementioned human back bony key point detection device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0115] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 9As shown, the computer device includes a processor, memory, and network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores preset neural network models and other data. The network interface communicates with external terminals via a network connection. When executed by the processor, the computer program implements a method for constructing a human back bony key point detection model or a method for detecting human back bony key points.

[0116] Those skilled in the art will understand that Figure 9 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0117] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0118] Obtain three-dimensional topographic data of the human back sample;

[0119] Preprocess the three-dimensional topographic data of the human back sample to obtain the sample normal rendering map;

[0120] Obtain the locations of the marked bony key points in the sample normal rendering image;

[0121] Based on the sample normal rendering map and the labeled bony key point locations, and with accuracy as the optimization objective, an initial neural network model is trained to obtain a human back bony key point detection model.

[0122] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0123] By using a 3D depth sensing component, the imaging area on the back of multiple sample objects is scanned to obtain 3D morphological data of the human back of the sample.

[0124] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0125] Principal component analysis was used to correct the orientation of the three-dimensional topographic data of the human back sample, and a normal rendering map of the sample was generated.

[0126] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0127] Principal component analysis was used to correct the orientation of the three-dimensional topographic data of the human back sample, generating an initial sample normal rendering map. The back region was then cropped and data augmented sequentially on the initial sample normal rendering map to obtain the final sample normal rendering map.

[0128] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0129] Push the sample normal rendering image to a third-party expert annotation platform; receive the sample normal rendering image with annotated bony key points from the third-party expert annotation platform, and obtain the location of the annotated bony key points in the sample normal rendering image.

[0130] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0131] Acquire three-dimensional topographic data of the human back;

[0132] Normal rendering is performed on the three-dimensional topographic data of the human back to obtain a normal rendering map.

[0133] The normal rendering map is input into the human back bony key point detection model to obtain the bony key point positions in the normal rendering map. The human back bony key point detection model is constructed by the above method.

[0134] Based on the location of bony key points in the normal rendering image, the location of bony key points on the human back can be obtained.

[0135] In one embodiment, the processor, when executing a computer program, also performs the following steps:

[0136] Obtain the coordinate correspondence between pixel coordinates in the normal rendering image and the coordinates of the three-dimensional shape data of the human back; based on the location of bony key points in the normal rendering image and the coordinate correspondence, obtain the three-dimensional coordinates of the bony key points of the human back.

[0137] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0138] Obtain three-dimensional topographic data of the human back sample;

[0139] Preprocess the three-dimensional topographic data of the human back sample to obtain the sample normal rendering map;

[0140] Obtain the locations of the marked bony key points in the sample normal rendering image;

[0141] Based on the sample normal rendering map and the labeled bony key point locations, and with accuracy as the optimization objective, an initial neural network model is trained to obtain a human back bony key point detection model.

[0142] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0143] By using a 3D depth sensing component, the imaging area on the back of multiple sample objects is scanned to obtain 3D morphological data of the human back of the sample.

[0144] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0145] Principal component analysis was used to correct the orientation of the three-dimensional topographic data of the human back sample, and a normal rendering map of the sample was generated.

[0146] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0147] Principal component analysis was used to correct the orientation of the three-dimensional topographic data of the human back sample, generating an initial sample normal rendering map. The back region was then cropped and data augmented sequentially on the initial sample normal rendering map to obtain the final sample normal rendering map.

[0148] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0149] Push the sample normal rendering image to a third-party expert annotation platform; receive the sample normal rendering image with annotated bony key points from the third-party expert annotation platform, and obtain the location of the annotated bony key points in the sample normal rendering image.

[0150] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0151] Acquire three-dimensional topographic data of the human back;

[0152] Normal rendering is performed on the three-dimensional topographic data of the human back to obtain a normal rendering map.

[0153] The normal rendering map is input into the human back bony key point detection model to obtain the bony key point positions in the normal rendering map. The human back bony key point detection model is constructed by the above method.

[0154] Based on the location of bony key points in the normal rendering image, the location of bony key points on the human back can be obtained.

[0155] In one embodiment, when the computer program is executed by a processor, it also performs the following steps:

[0156] Obtain the coordinate correspondence between pixel coordinates in the normal rendering image and the coordinates of the three-dimensional shape data of the human back; based on the location of bony key points in the normal rendering image and the coordinate correspondence, obtain the three-dimensional coordinates of the bony key points of the human back.

[0157] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0158] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0159] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A method for constructing a human back bone key point detection model, characterized in that, The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; obtaining the labeled skeletal key point positions in the sample normal rendering graph; training an initial neural network model according to the sample normal rendering graph and the labeled skeletal key point positions, and taking accuracy as an optimization target to obtain a human back skeletal key point detection model.

2. The method of claim 1, wherein, The method comprises: scanning the back imaging area of a plurality of sample objects by a three-dimensional depth sensing component to obtain sample human back three-dimensional topography data.

3. The method of claim 1, wherein, The preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph comprises: correcting the direction of the sample human back three-dimensional topography data by principal component analysis to generate a sample normal rendering graph.

4. The method of claim 3, wherein, The preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph comprises: correcting the direction of the sample human back three-dimensional topography data by principal component analysis to generate an initial sample normal rendering graph; sequentially performing back region cropping and data enhancement on the initial sample normal rendering graph to obtain a sample normal rendering graph.

5. The method of claim 1, wherein, The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; 6.A method for detecting key points of human back bones, characterized in that, The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; 7. The method of claim 6, wherein, preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; 8.A human back bone key point detection model construction device, characterized in that, preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back three-dimensional topography data to obtain a sample normal rendering graph; The method comprises: obtaining sample human back three-dimensional topography data; preprocessing the sample human back 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-8 when the computer program is executed by the processor. When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images