Hand acupuncture point detection method and system for learning attention based on region correlation coefficient

Abstract

The invention relates to the technical field of image processing, in particular to a hand acupuncture point detection method and system based on region correlation coefficient learning attention, and the method comprises the steps: obtaining a hand image, and carrying out the feature extraction of the hand image, and obtaining a first feature map; processing the first feature map by using a regional correlation coefficient learning attention network to obtain a third feature map; wherein the regional correlation coefficient learning attention network comprises a self-adaptive block perception attention module and a self-adaptive correlation coefficient channel attention module, and fine-grained acupuncture points in the first feature map are detected through the self-adaptive block perception attention module to obtain a second feature map; modeling acupoint features of different skin colors in the second feature map through an adaptive correlation coefficient channel attention module to obtain a third feature map; predicting labels and coordinate points of hand acupuncture points based on the third feature map; according to the invention, fine-grained acupuncture points can be detected, and acupuncture points under different skin colors can be positioned, so that the effectiveness of hand acupuncture point detection is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and particularly to a method and system for detecting hand acupoints based on region correlation coefficient learning attention. Background Art

[0002] Traditional Chinese acupuncture (CA) is a unique treatment method in traditional Chinese medicine, mainly by stimulating acupoints to regulate the flow of qi (energy) and blood to treat various conditions such as headache, insomnia, and facial paralysis.

[0003] Traditional Chinese acupuncture (CA) plays an important role in the treatment, prevention, and health care of diseases. However, due to the complexity of the task, it is difficult to detect hand acupoints from the body surface, mainly manifested as: 1) the acupoint area is extremely similar to the surrounding parts of the body surface; 2) the skin color depth of the hand is different. With the development of computer hardware technology, deep neural networks have been widely applied in various fields. Due to their powerful feature recognition ability, deep neural networks have also been applied in traditional Chinese acupuncture.

[0004] Therefore, it is necessary to design computer-aided tools to train acupuncturists or assist doctors in acupuncture work. Combining deep neural networks to solve the above technical problems can detect fine-grained acupoints and locate acupoints under different skin colors, thereby improving the effectiveness of hand acupoint detection. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for detecting hand acupoints based on region correlation coefficient learning attention, which can detect fine-grained acupoints and locate acupoints under different skin colors, thereby improving the effectiveness of hand acupoint detection.

[0006] To achieve the above purpose, the present invention provides the following technical solutions: In the first aspect, an embodiment of the present invention provides a method for detecting hand acupoints based on region correlation coefficient learning attention, the method comprising the following steps: Obtain a hand image, perform feature extraction on the hand image to obtain a first feature map; Process the first feature map using a region correlation coefficient learning attention network to obtain a third feature map; wherein, the region correlation coefficient learning attention network includes an adaptive block perception attention module and an adaptive correlation coefficient channel attention module, detect fine-grained acupoints in the first feature map through the adaptive block perception attention module to obtain a second feature map; model the acupoint features of different skin colors in the second feature map through the adaptive correlation coefficient channel attention module to obtain a third feature map; Predict the labels and coordinate points of hand acupoints based on the third feature map.

[0007] Preferably, extracting features from the hand image to obtain a first feature map includes: Extracting a first feature map from the hand image using a head convolution.

[0008] Preferably, detecting fine-grained acupoints in the first feature map through the adaptive block-aware attention module to obtain a second feature map includes: Evenly dividing the first feature map into multiple patches, and connecting the multiple patches to generate a patch feature map; Processing the patch feature map using an average criterion and inputting it into a patch attention module to output attention weights; Processing the attention weights through an adaptive mask to obtain a second feature map.

[0009] Preferably, the expression of the attention weights is: K = Softmax{MLP{Norm[MLP(J)]}}; Where K represents the attention weights, Softmax{·} represents performing the Softmax activation function processing, MLP{·} represents performing the multi-layer perceptron processing, and Norm{·} represents performing the normalization processing.

[0010] Preferably, processing the attention weights through an adaptive mask to obtain a second feature map includes: Normalizing the attention weights and then multiplying them by learnable parameters to obtain a mask map; Multiplying the attention weights by the mask map to obtain an attention feature map, and reshaping the attention feature map into a second feature map.

[0011] Preferably, modeling acupoint features of different skin colors in the second feature map through the adaptive correlation coefficient channel attention module to obtain a third feature map includes: Inputting the second feature map into a channel correlation selector to generate a channel selection mask and an inverse mask; Multiplying the second feature map by the mask to obtain an effective feature map; multiplying the second feature map by the inverse mask to obtain a redundant feature map; Multiplying a learnable factor by the second feature map and then adding it to the effective feature map and the redundant feature map to obtain a third feature map.

[0012] Preferably, inputting the second feature map into a channel correlation selector to generate a channel selection mask and an inverse mask includes: Performing pooling processing, convolution processing, and activation processing on the second feature map in sequence to generate a channel selection mask and an inverse mask respectively.

[0013] Preferably, predicting the positions and labels of hand acupoints based on the third feature map includes: Performing an inverted bottleneck process and a deconvolution process on the third feature map to obtain the coordinate point positions and labels of hand acupoints.

[0014] In a second aspect, an embodiment of the present invention provides a hand acupoint detection system based on region - related coefficient learning attention, and the system includes: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method as described in any one of the above.

[0015] The beneficial effects of the present invention are as follows: Through the adaptive block - aware attention module, the present invention can capture detailed information at multiple scales and is designed to detect fine - grained acupoints; through the adaptive correlation - coefficient channel attention module, channel correlation can be modeled to design the positioning of hand acupoints on various skin tones. The present invention can detect fine - grained acupoints and locate acupoints under different skin tones. The present invention can detect fine - grained acupoints and locate acupoints under different skin tones, thereby improving the effectiveness of hand acupoint detection. Description of the Drawings

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or in the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0017] Figure 1 is a schematic flowchart of a method for detecting hand acupoints based on region - related coefficient learning attention in an embodiment of the present invention; Figure 2 is a framework diagram of a region - related coefficient learning attention network in an embodiment of the present invention; Figure 3 In [reference figure], a is a comparison schematic diagram of the appearance between the acupoint area and the surrounding area in an embodiment of the present invention; b is another comparison schematic diagram of the appearance between the acupoint area and the surrounding area in an embodiment of the present invention; c is a comparison schematic diagram of the skin tone between the acupoint area and the surrounding area in an embodiment of the present invention; d is another comparison schematic diagram of the skin tone between the acupoint area and the surrounding area in an embodiment of the present invention; Figure 4In it, a is a schematic diagram of 11 acupoints selected from the palm in the embodiment of the present invention; b is a schematic diagram of 7 acupoints selected from the back of the hand in the embodiment of the present invention; c is a schematic diagram of the reference distance of the palm in the embodiment of the present invention; d is a schematic diagram of the reference distance of the back of the hand in the embodiment of the present invention; Figure 5 It is a schematic structural diagram of a hand acupoint detection system based on region correlation coefficient learning attention in the embodiment of the present invention. Detailed implementation manners

[0018] The following will clearly and completely describe the concept, specific structure and technical effects generated by the present invention in combination with embodiments and drawings, so as to fully understand the purpose, scheme and effects of the present invention. It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0019] In the related art, a new method for acupoint positioning that combines RGB and depth images to guide robot-controlled acupuncture is disclosed. However, both RGB and depth images are easily affected by skin color changes, which will reduce the image quality and the accuracy of acupoint positioning. In addition, a hand acupoint positioning method that combines a dual attention mechanism with a cascaded network model is disclosed, which ensures real-time performance and robustness under complex conditions, including illumination changes. However, the changes in skin texture and color often lead to the acupoint area being very similar to the surrounding area, resulting in difficulties in hand acupoint positioning.

[0020] A facial acupoint detection method based on deep learning of images is disclosed, which uses a high-resolution network for feature extraction. However, the change in light intensity still affects the image quality and feature extraction.

[0021] Two network-based methods are disclosed, which use computer vision to automatically identify the acupoints on the face and hand in real time, improving the accuracy and accessibility of acupuncture. However, these methods show limited robustness in distinguishing similar acupoints.

[0022] In summary, although these deep learning-based models have shown good performance, there are still challenges in Chinese acupuncture. First, the acupoint area is very similar to the surrounding skin area, and it is difficult to distinguish them only based on appearance. Subtle differences are often difficult to detect, resulting in potential misidentifications, especially in areas with low contrast. Second, the change in skin color complicates the observation or palpation of specific signs on the hand acupoints. For example, a darker skin color may make it more difficult to identify details or shallow depressions, thus increasing the challenge of acupoint positioning.

[0023] To solve the above technical problems, the present invention proposes a Region Correlation Learning Attention Network (RCLA-Net) for detecting acupoints on the hand in Chinese acupuncture, which includes an Adaptive Patch Aware Attention (APAA) module and an Adaptive Correlation Channel Attention (ACCA) module; the fine-grained acupoints are detected by capturing the feature information of small patches through the Adaptive Patch Aware Attention module, and the acupoints under different skin colors are located by channel correlation modeling through the Adaptive Correlation Channel Attention module, thereby improving the effectiveness of hand acupoint detection.

[0024] Referring to Figure 1 , the present invention provides a hand acupoint detection method based on region correlation coefficient learning attention, and the method includes the following steps: S100, obtain a hand image, perform feature extraction on the hand image to obtain a first feature map; S200, process the first feature map by using a region correlation coefficient learning attention network to obtain a third feature map; wherein, the region correlation coefficient learning attention network includes an Adaptive Patch Aware Attention module and an Adaptive Correlation Channel Attention module, and the fine-grained acupoints in the first feature map are detected through the Adaptive Patch Aware Attention module to obtain a second feature map; the acupoint features of different skin colors in the second feature map are modeled through the Adaptive Correlation Channel Attention module to obtain a third feature map; S300, predict the labels and coordinate points of hand acupoints based on the third feature map.

[0025] Through the Adaptive Patch Aware Attention module, the present invention can capture detailed information at multiple scales and is designed to detect fine-grained acupoints; through the Adaptive Correlation Channel Attention module, the present invention can model channel correlation and design to locate hand acupoints on various skin colors; the present invention can detect fine-grained acupoints and locate acupoints under different skin colors, thereby improving the effectiveness of hand acupoint detection.

[0026] In some improved embodiments, in S100, the performing feature extraction on the hand image to obtain a first feature map includes: Extract a first feature map from the hand image using a head convolution.

[0027] In some improved embodiments, in S200, the detecting the fine-grained acupoints in the first feature map through the Adaptive Patch Aware Attention module to obtain a second feature map includes: The first feature map is evenly divided into multiple patches, and the multiple patches are connected to generate a patch feature map; The patch feature map is processed using an average criterion and then input into a patch attention module to output attention weights; The attention weights are processed through an adaptive mask to obtain a second feature map.

[0028] In some improved embodiments, the expression of the attention weights is: K = Softmax{MLP{Norm[MLP(J)]}}; where K represents the attention weights, Softmax{·} represents the processing of the Softmax activation function, MLP{·} represents the processing of a multi-layer perceptron, and Norm{·} represents the processing of normalization.

[0029] In some improved embodiments, the processing of the attention weights through an adaptive mask to obtain a second feature map includes: After normalizing the attention weights, they are multiplied by learnable parameters to obtain a mask map; The attention weights are multiplied by the mask map to obtain an attention feature map, and the attention feature map is reshaped into a second feature map.

[0030] As Figure 2 shown, the Region-Correlation Coefficient Learning Attention Network (RCLA-Net) includes two key parts: an Adaptive Patch-Aware Attention (APAA) module and an Adaptive Correlation Coefficient Channel Attention (ACCA) module. The detailed description of each module is as follows.

[0031] The processing procedure of the Adaptive Patch-Aware Attention module for fine acupoint recognition is as follows: In traditional Chinese acupuncture, the acupoint area looks very similar to the area around the hand surface. First, both areas exhibit consistent skin texture without obvious boundaries that distinguish natural skin patterns. For example, the texture around the acupoints on the hand naturally extends to the acupoints themselves. In addition, surface landmarks such as muscle contours and joint positions are equally aligned in both the acupoint area and the surrounding area, making it visually difficult to distinguish, as Figure 3 shown in a and b of. Based on these observations, the present invention designs an Adaptive Patch-Aware Attention (APAA) module to perform acupoint recognition by capturing fine-grained features on small patches, as Figure 2 shown in the top module of.

[0032] The expression definition of the Adaptive Patch-Aware Attention module is: APPA(I) = AM[PA(I)]; Among them, PA and AM represent patch attention and adaptive mask respectively, which can be described as follows: First, the first feature map I is evenly divided into small patches, and then they are concatenated to generate a patch feature map J, which is then processed with an average criterion.

[0033] As Figure 2 As shown in the adaptive patch-aware attention module at the top, the first feature map is divided into small patches, enabling attention to focus on local details at different positions. This allows the adaptive patch-aware attention module to analyze each local area in a fine-grained manner, further differentiating the acupoint area from the surrounding similar areas.

[0034] Then, the patch feature map J is input into a multi-layer perceptron (MLP) to generate attention weights K, defined as: PA(J)=Softmax{MLP{Norm[MLP(J)]}}; where PA(J) represents performing patch attention processing on the patch feature map J, and the attention weights K = PA(J).

[0035] Next, the attention weights K are processed through an adaptive mask (AM). Specifically, the attention weights K are normalized and multiplied by a learnable parameter α to obtain a mask map M.

[0036] Finally, the attention weights K are multiplied by the mask map M to obtain an attention feature map L, which is then reshaped into a second feature map X.

[0037] In some improved embodiments, in S200, the step of modeling acupoint features of different skin colors in the second feature map through the adaptive correlation coefficient channel attention module to obtain a third feature map includes: Inputting the second feature map into a channel correlation selector to generate a channel selection mask and an inverse mask; Multiplying the second feature map by the mask to obtain an effective feature map; multiplying the second feature map by the inverse mask to obtain a redundant feature map; Multiplying a learnable factor by the second feature map and then adding it to the effective feature map and the redundant feature map to obtain a third feature map.

[0038] In some improved embodiments, the step of inputting the second feature map into a channel correlation selector to generate a channel selection mask and an inverse mask includes: Performing pooling processing, convolutional processing, and activation processing on the second feature map in sequence to generate a channel selection mask and an inverse mask respectively.

[0039] Specifically, the convolutional kernel size of the convolutional processing is 1×1; the activation function of the activation processing is the ReLU function.

[0040] The processing procedure of the adaptive correlation coefficient channel attention module for skin color modeling is introduced in detail as follows: In acupuncture, skin color is crucial for determining the clarity, color accuracy, depth, and three-dimensional effect of acupoint imaging on the body surface. Specifically, excessive skin color will obscure details, while light skin color will cause changes in imaging, as shown in c and d of Figure 3 . Therefore, accurate skin color modeling is crucial for optimal visualization and accurate assessment of the acupoint area. To solve this problem, the present invention designs an adaptive correlation coefficient channel attention (ACCA) module to model acupoint features under different skin colors by modeling channel correlation, as shown in the adaptive correlation coefficient channel attention module at the bottom. The expression of the adaptive correlation coefficient channel attention module is defined as: Figure 2 ; ; where H(·) is the channel correlation selector, X is the second feature map, β is the learnable factor, and Z is the third feature map.

[0041] The processing procedure of the adaptive correlation coefficient channel attention module for the second feature map X is described as follows: First, input the second feature map X into the channel correlation selector H(·). The channel correlation selector H(·) includes pooling, convolution, and ReLU blocks to generate a channel selection mask H(X) and an inverse mask 1 - H(X).

[0042] Next, multiply the second feature map X by the mask H(X) and the inverse mask 1 - H(X) respectively to obtain an effective feature map Y and a redundant feature map Yˊ, and the expressions are: ; Finally, multiply the learnable factor β by the second feature map X, and then add it to the effective feature map Y and the redundant feature map Yˊ to obtain the third feature map Z.

[0043] Based on the above observations, the adaptive correlation coefficient channel attention module allows the regional correlation coefficient learning attention network to focus on effective features while suppressing redundant information, thereby improving the overall performance of acupoint localization for different skin colors.

[0044] In some improved embodiments, in S300, predicting the positions and labels of hand acupoints based on the third feature map includes: Performing an inverted bottleneck process and a deconvolution process on the third feature map to obtain the coordinate point positions and labels of hand acupoints.

[0045] To verify the effectiveness of the proposed regional correlation coefficient learning attention network for hand acupoint detection, the present invention has conducted extensive experiments on a hand acupoint dataset.

[0046] A. Hand Acupoint Dataset: The 11K Hand Subset (11KHS) dataset is adopted. This dataset consists of 3,860 palm images and 4,175 dorsal hand images, with a resolution of 1,600×1,200 for each image, and is collected from subjects aged 18 - 75. Specifically, 11 acupoints for traditional Chinese medicine acupuncture are selected from the palm area, namely: Sishengxue (Extraordinary Point EX - UE10, a total of four acupoints), Shixuanxue (Extraordinary Point EX - UE11, a total of four acupoints), Shaofu (Heart Meridian HT8), Laogong (Pericardium Meridian PC8), and Zhongchong (Pericardium Meridian PC9), as shown in Figure 4 a in. In addition, 7 acupoints for traditional Chinese medicine acupuncture are selected from the dorsal hand area, namely: Hegu (Large Intestine Meridian LI4), Zhongzhu (Triple Energizer Meridian SJ3), Yemen (Triple Energizer Meridian SJ2), Shaoze (Small Intestine Meridian SI1), Shaochong (Heart Meridian HT9), Guanchong (Triple Energizer Meridian SJ1), and Shangyang (Large Intestine Meridian LI1), as shown in Figure 4 b in.

[0047] B. Evaluation Metrics: To evaluate the performance of the region - related coefficient learning attention network, the present invention regards the acupoint detection rate (ADR) as the evaluation metric: (5); In the formula, N is the total number of hand acupoints, and C is the number of correctly predicted acupoints. When the error E is less than the threshold λ (which is set to 0.04, 0.06, 0.10, 0.12 respectively in the present invention), the predicted acupoint is considered a correct acupoint, and E is defined as: (6); where and are the j - th predicted acupoint and the true label respectively.

[0048] In addition, to eliminate the influence of image size and the size of the hand itself on the results, the distance between two fixed points is used as the reference value for normalization. Specifically, for palm images, the Euclidean distance between Laogong and Shaofu is set as the normalization factor. For dorsal hand images, the Euclidean distance between Yemen and Zhongzhu is used as the normalization factor, as shown in Figure 4 c and d in.

[0049] Comparison with hand acupoint detection methods on the 11KHS dataset; In this invention, on the 11KHS dataset, the RCLA-Net (Region Correlation Coefficient Learning Attention Network) is compared with five hand acupoint detection methods, including SimpleNet, EfficientHRNet, HigherHRNet, LECANet, and LitePose. The quantitative results are shown in Table 1.

[0050] Table 1: Quantitative Results on the 11KHS Dataset

[0051] As can be seen from Table 1, the performance of the proposed Region Correlation Coefficient Learning Attention Network (RCLA-Net) is better than other methods. When the error threshold λ is set to 0.04, 0.06, 0.10, and 0.12, the ADRs on the 11KHS dataset are 31.24%, 49.47%, 69.28%, and 76.11% respectively. In addition, the Region Correlation Coefficient Learning Attention Network proposed in this invention improves the second-best method LitePose. The results show that in hand acupoint detection, the adaptive patch correlation attention mechanism is more effective than multi-branch information fusion. In addition, RCLA-Net is significantly better than SimpleNet because SimpleNet is difficult to effectively learn fine-grained local features. It can be seen that the experiments on the benchmark dataset prove that the Region Correlation Coefficient Learning Attention Network proposed in this invention can improve the effectiveness of Chinese acupuncture hand acupoint detection.

[0052] Corresponding to Figure 1 the method of, referring to Figure 5 this, an embodiment of the present invention provides a hand acupoint detection system based on region correlation coefficient learning attention, including: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the above method.

[0053] It can be seen that the content in the above method embodiments is applicable to the system embodiments of this system. The functions specifically implemented by the system embodiments of this system are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.

[0054] In addition, an embodiment of the present invention also discloses a computer program product or a computer program, which is stored in a computer-readable storage medium. The processor of the computer device can read the computer program from the computer-readable storage medium, and the processor executes the computer program, so that the computer device executes the above-mentioned method. Similarly, the content in the above method embodiments is applicable to this storage medium embodiment. The functions specifically implemented by this storage medium embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those of the above method embodiments.

[0055] Those of ordinary skill in the art will appreciate that all or some of the methods and systems disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. Some physical components or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or transitory medium). As is well known to those of ordinary skill in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassette, tape, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, as is well known to those of ordinary skill in the art, communication media typically includes computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transmission mechanism, and may include any information delivery medium.

[0056] The above is a specific description of the preferred embodiments of the present disclosure, but the present disclosure is not limited to the above embodiments. Those skilled in the art can also make various equivalent deformations or substitutions without departing from the spirit of the present disclosure, and these equivalent deformations or substitutions are all included within the scope defined by the claims of the present disclosure.

Claims

1. A hand acupoint detection method based on learning attention with regional correlation coefficient, characterized in that The method includes the following steps: Obtain a hand image, extract features from the hand image to obtain a first feature map; Process the first feature map using a region correlation coefficient learning attention network to obtain a third feature map; wherein, the region correlation coefficient learning attention network includes an adaptive block perception attention module and an adaptive correlation coefficient channel attention module, detect fine-grained acupoints in the first feature map through the adaptive block perception attention module to obtain a second feature map; model the acupoint features of different skin colors in the second feature map through the adaptive correlation coefficient channel attention module to obtain a third feature map; Predict the labels and coordinate points of hand acupoints based on the third feature map.

2. The method according to claim 1, characterized in that, The extracting features from the hand image to obtain a first feature map includes: Extract a first feature map from the hand image using a head convolution.

3. The method according to claim 1, characterized in that The detecting fine-grained acupoints in the first feature map through the adaptive block perception attention module to obtain a second feature map includes: Evenly divide the first feature map into multiple patches, and connect the multiple patches to generate a patch feature map; Process the patch feature map using an average criterion and input it into a patch attention module to output attention weights; Process the attention weights through an adaptive mask to obtain a second feature map.

4. The method according to claim 3, characterized in that The expression of the attention weights is: K = Softmax{MLP{Norm[MLP(J)]}}; wherein, K represents the attention weights, Softmax{·} represents performing Softmax activation function processing, MLP{·} represents performing multi-layer perceptron processing, and Norm{·} represents performing normalization processing.

5. The method according to claim 3, characterized in that The processing the attention weights through an adaptive mask to obtain a second feature map includes: After normalizing the attention weights, multiply them by learnable parameters to obtain a mask map; Multiply the attention weights by the mask map to obtain an attention feature map, and reshape the attention feature map into a second feature map.

6. The method according to claim 1, characterized in that, The modeling the acupoint features of different skin colors in the second feature map through the adaptive correlation coefficient channel attention module to obtain a third feature map includes: Input the second feature map into a channel correlation selector to generate a channel selection mask and an inverse mask; Multiply the second feature map by the mask to obtain an effective feature map; multiply the second feature map by the inverse mask to obtain a redundant feature map; Multiply a learnable factor by the second feature map and add it to the effective feature map and the redundant feature map to obtain a third feature map.

7. The method according to claim 6, wherein The inputting the second feature map into a channel correlation selector to generate a channel selection mask and an inverse mask includes: Perform pooling processing, convolution processing, and activation processing on the second feature map in sequence to generate a channel selection mask and an inverse mask respectively.

8. The method according to claim 1, characterized in that, The predicting the positions and labels of hand acupoints based on the third feature map includes: Perform an inverted bottleneck process and a deconvolution process on the third feature map to obtain the coordinate point positions and labels of hand acupoints.

9. A hand acupoint detection system based on learning attention with regional correlation coefficient, characterized in that, The system includes: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor such that the at least one processor implements the method according to any one of claims 1 to 8.

Images