An operation optimization method for a lower limb hip-knee rehabilitation device based on machine vision

Abstract

The present invention discloses an operation optimization method for a lower limb hip-knee rehabilitation device based on machine vision. First, multi-angle video data of a rehabilitation person using the device is acquired, key frames are extracted and the target object is located, and the movement characteristics are analyzed and the movement cycle is divided; the HOG algorithm is used to detect, segment and extract movement characteristics, and multi-scale fusion is performed through a Transformer model to generate fused movement characteristic data; based on the rehabilitation plan, a GAN generation model is used to simulate the standard movement characteristics in an ideal state, and multi-scale fusion is performed to generate standard fused characteristic data; the similarity between the fused movement characteristic data and the standard fused characteristic data is calculated, the operation state of the rehabilitation device is evaluated and a rehabilitation movement index is generated; based on the rehabilitation movement index, the parameters of the device module are optimized. Through the present invention, the coherence, standardness and corresponding rehabilitation state of the operation of the rehabilitation device can be accurately analyzed, the rehabilitation effect can be effectively evaluated, and the dynamic optimization and real-time adjustment of the device parameters can be realized.

Description

Technical Field

[0001] The present invention relates to the field of operation analysis of rehabilitation equipment, and more specifically, to a method for optimizing the operation of a lower limb hip-knee rehabilitation equipment based on machine vision. Background Art

[0002] For the elderly, the treatment of lower limb hip-knee joint diseases is particularly important. Especially for the movement rehabilitation of the lower limbs, which is crucial for people's ability to move freely, it becomes even more prominent. At the same time, problems such as trauma and disability will also lead to movement dysfunction, further increasing people's demand for rehabilitation training. The traditional method of lower limb hip-knee rehabilitation generally uses manual methods to achieve the rehabilitation movement of patients. This method is inefficient and has poor results. For general rehabilitation equipment, the prior art lacks effective monitoring and effectiveness evaluation of the rehabilitation movement state, and lacks refined analysis of the equipment operation effect during the rehabilitation movement process, resulting in the equipment being difficult to adapt to different rehabilitation personnel and difficult to meet the target rehabilitation plan. Summary of the Invention

[0003] The present invention overcomes the defects of the prior art and proposes a method for optimizing the operation of a lower limb hip-knee rehabilitation equipment based on machine vision.

[0004] The first aspect of the present invention provides a method for optimizing the operation of a lower limb hip-knee rehabilitation equipment based on machine vision, including:

[0005] Obtaining multi-angle video data of a rehabilitation person when using the lower limb hip-knee rehabilitation equipment through machine vision technology;

[0006] Based on each angle, extracting key frames from the video data to form a rehabilitation process image data set. Each angle corresponds to an image data set. Locating the target object of the rehabilitation equipment in the image data set, analyzing the movement characteristics of the target object, and combining the rehabilitation plan, dividing the usage process time period into multiple motion cycles;

[0007] Within a motion cycle, performing moving target detection, target segmentation, and feature extraction on the image data sets of multiple angles based on the HOG algorithm to form motion feature data for each image data set. Taking one angle as one scale, performing multi-scale fusion on the motion feature data through a hierarchical multi-scale Transformer model to generate fused motion feature data;

[0008] Based on the rehabilitation plan and the motion cycle, setting the multi-angle motion parameters of the moving target in the ideal state. Through the GAN generation model, combining the motion parameters to generate the simulated features of the moving target, obtaining the multi-angle standard motion features, and performing multi-scale fusion on the standard motion features through the hierarchical multi-scale Transformer model to generate standard fused feature data;

[0009] Take the fused motion feature data and the standard fusion feature data as two samples, calculate the similarity between the samples based on the Mahalanobis distance, and evaluate the operating state of the rehabilitation device based on the similarity, generating a rehabilitation motion index;

[0010] Calculate the corresponding rehabilitation motion index for each motion cycle, and compare the rehabilitation motion index with a preset range to optimize the settings of each module of the corresponding rehabilitation device.

[0011] In this solution, through machine vision technology, multi-angle video data of a rehabilitation person using a lower limb hip-knee rehabilitation device is obtained, specifically:

[0012] Through machine vision technology, set a multi-angle video acquisition device and apply it within the preset motion range of the rehabilitation person;

[0013] Based on the multi-angle video acquisition device, within the usage time period of a lower limb hip-knee rehabilitation device, obtain the corresponding multi-angle video data.

[0014] In this solution, based on each angle, extract key frames from the video data to form a rehabilitation process image dataset. Each angle corresponds to an image dataset. Locate the target object of the rehabilitation device for the image dataset, analyze the movement characteristics of the target object, and combine the rehabilitation plan to divide the usage process time period into multiple motion cycles, specifically:

[0015] Based on each angle, extract key frames from the video data to form a rehabilitation process image dataset;

[0016] Perform image noise reduction, enhancement, and normalization preprocessing on the image dataset;

[0017] Based on the rehabilitation device, set target objects for multiple device joint positions and store the key image features of the target objects;

[0018] Select an image dataset of one angle, and through the recognition module, perform recognition on multiple frames of the image dataset. The recognition process is to detect and locate the target object by combining the key image features, analyze the movement characteristics of the target object in multiple frames of images, and generate the movement data of the target object;

[0019] Analyze the movement data, combine the rehabilitation plan to perform coherence analysis on the actions of the rehabilitation person, and divide the usage process time period into multiple motion cycles.

[0020] In this solution, within one motion cycle, motion target detection, target segmentation, and feature extraction based on the HOG algorithm are performed on image datasets at multiple angles to form motion feature data for each image dataset. Taking one angle as one scale, multi-scale fusion of the motion feature data is performed through a hierarchical multi-scale Transformer model to generate fused motion feature data. Specifically:

[0021] The image dataset is grayscale processed. Based on the HOG algorithm, the gradient magnitude and direction of local pixel points are calculated for each frame of the image, and multiple object contour features are extracted. Target detection and segmentation are performed from the multiple object contour features to construct the target feature vector of each frame of the image;

[0022] Feature change analysis is performed on the target feature vectors of two adjacent frames of images to generate feature difference information. The target feature vector of each frame of the image and the feature difference information are integrated to form motion feature data;

[0023] For each image dataset, corresponding motion feature data is generated;

[0024] Through the hierarchical multi-scale Transformer model, taking each angle as one scale, multiple Transformer sub-modules are constructed. The multiple motion feature data are respectively input into the Transformer sub-modules for feature transformation to form multi-scale feature data;

[0025] Through the attention mechanism, the multi-scale feature data are fused and represented to generate fused motion feature data.

[0026] In this solution, each motion cycle includes corresponding fused motion feature data.

[0027] In this solution, based on the rehabilitation plan and the motion cycle, the motion parameters of the motion target at multiple angles in the ideal state are set. Through the GAN generation model, combined with the motion parameters, the simulated features of the motion target are generated to obtain the standard motion features at multiple angles. The standard motion features are multi-scale fused through the hierarchical multi-scale Transformer model to generate standard fused feature data. Specifically:

[0028] Based on the rehabilitation plan and the division of the motion cycle, the motion parameters of the motion target at multiple angles in the ideal state are set;

[0029] According to the motion parameters, multiple angle image frames of the motion target in the initial state and the motion state are obtained from the system database. For each angle of the image frames, they are marked as the initial image set;

[0030] Extract features from the initial image set based on the HOG algorithm to form initial motion feature data;

[0031] Construct a GAN-based generation model, import the initial motion feature data and motion parameters into the generation model for training and learning, cyclically generate simulated motion feature data. In each training process, analyze the object motion state based on the simulated motion feature data and the initial motion feature data, use the consistency between the object motion state and the motion parameters as the accuracy rate for judging the generated data, and perform cyclic training until the generation model reaches the preset number of training times;

[0032] Generate standard motion features through the trained generation model;

[0033] Generate standard motion features from multiple angles based on the initial image set from multiple angles;

[0034] Perform multi-scale fusion on the standard motion features through a hierarchical multi-scale Transformer model to generate standard fusion feature data.

[0035] In this solution, use the fused motion feature data and the standard fusion feature data as two samples, calculate the similarity between the samples based on the Mahalanobis distance, and evaluate the operating state of the rehabilitation device based on the similarity to generate a rehabilitation motion index. Specifically:

[0036] Use the fused motion feature data and the standard fusion feature data as two samples, and calculate the distance value between the samples based on the Mahalanobis distance;

[0037] Analyze the similarity between the samples through the distance value, and evaluate the operating state of the rehabilitation device based on the similarity to generate a rehabilitation motion index;

[0038] The rehabilitation motion index is proportional to the similarity.

[0039] In this solution, calculate the corresponding rehabilitation motion index for each motion cycle, and compare the rehabilitation motion index with the preset range to optimize the settings of each module of the corresponding rehabilitation device. Specifically:

[0040] Perform calculation and analysis for each motion cycle to obtain the corresponding rehabilitation motion index;

[0041] Judge whether the rehabilitation motion index of each motion cycle meets the preset range. If not, mark the corresponding motion cycle as an optimized motion cycle;

[0042] Mark the device modules operated by the rehabilitation device during the optimized motion cycle, optimize the parameter settings of the device modules, and generate an optimization plan.

[0043] In a second aspect of the present invention, there is also provided an operation optimization system for a lower limb hip-knee rehabilitation device based on machine vision. The system includes: a memory and a processor. The memory includes an operation optimization program for a lower limb hip-knee rehabilitation device based on machine vision. When the operation optimization program for a lower limb hip-knee rehabilitation device based on machine vision is executed by the processor, the following steps are implemented:

[0044] Through machine vision technology, multi-angle video data of a rehabilitation person using a lower limb hip-knee rehabilitation device is acquired;

[0045] Based on each angle, key frames are extracted from the video data to form a rehabilitation process image data set. Each angle corresponds to an image data set. The target object of the rehabilitation device is located in the image data set, and the movement characteristics of the target object are analyzed. Combining with the rehabilitation plan, multiple movement cycles are divided in the usage process time period;

[0046] Within a movement cycle, motion target detection, target segmentation, and feature extraction based on the HOG algorithm are performed on the image data sets of multiple angles to form motion feature data for each image data set. Taking one angle as one scale, the motion feature data is multi-scale fused through a hierarchical multi-scale Transformer model to generate fused motion feature data;

[0047] Based on the rehabilitation plan and the movement cycle, multi-angle motion parameters of the motion target in an ideal state are set. Through the GAN generation model, combined with the motion parameters, simulated features of the motion target are generated to obtain multi-angle standard motion features. The standard motion features are multi-scale fused through a hierarchical multi-scale Transformer model to generate standard fused feature data;

[0048] The fused motion feature data and the standard fused feature data are used as two samples, and the similarity between the samples is calculated based on the Mahalanobis distance. Based on the similarity, the operation state of the rehabilitation device is evaluated to generate a rehabilitation motion index;

[0049] For each movement cycle, the corresponding rehabilitation motion index is calculated, and by comparing the rehabilitation motion index with a preset range, optimization settings are made for each module of the corresponding rehabilitation device.

[0050] In a third aspect of the present invention, there is also provided a computer-readable storage medium. The computer-readable storage medium includes an operation optimization program for a lower limb hip-knee rehabilitation device based on machine vision. When the operation optimization program for a lower limb hip-knee rehabilitation device based on machine vision is executed by a processor, the steps of the operation optimization method for a lower limb hip-knee rehabilitation device based on machine vision as described in any one of the above are implemented.

[0051] The present invention discloses an operation optimization method for a lower limb hip-knee rehabilitation device based on machine vision. First, multi-angle video data of a rehabilitation person using the device is acquired, key frames are extracted and the target object is located, and the movement characteristics are analyzed and the movement cycle is divided; the HOG algorithm is used to detect, segment and extract the movement characteristics, and multi-scale fusion is performed through a Transformer model to generate fused movement characteristic data; based on the rehabilitation plan, a GAN generation model is used to simulate the standard movement characteristics in an ideal state, and multi-scale fusion is performed to generate standard fused characteristic data; the similarity between the fused movement characteristic data and the standard fused characteristic data is calculated, the operation state of the rehabilitation device is evaluated, and a rehabilitation movement index is generated; based on the rehabilitation movement index, the parameters of the device module are optimized. Through the present invention, the coherence, standardness and corresponding rehabilitation state of the operation of the rehabilitation device can be accurately analyzed, the rehabilitation effect can be effectively evaluated, and the dynamic optimization and real-time adjustment of the device parameters can be realized. Brief Description of the Drawings

[0052] Figure 1 The flowchart of an operation optimization method for a lower limb hip-knee rehabilitation device based on machine vision according to the present invention is shown;

[0053] Figure 2 The block diagram of an operation optimization system for a lower limb hip-knee rehabilitation device based on machine vision according to the present invention is shown. Detailed Description of the Embodiments

[0054] In order to more clearly understand the above objects, features and advantages of the present invention, the present invention will be further described in detail below with reference to the drawings and specific embodiments. It should be noted that, without conflict, the embodiments of the present application and the features in the embodiments may be combined with each other.

[0055] Many specific details are set forth in the following description in order to fully understand the present invention. However, the present invention may be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited by the specific embodiments disclosed below.

[0056] Figure 1 The flowchart of an operation optimization method for a lower limb hip-knee rehabilitation device based on machine vision according to the present invention is shown.

[0057] As Figure 1 shown, a first aspect of the present invention provides an operation optimization method for a lower limb hip-knee rehabilitation device based on machine vision, including:

[0058] S102, through machine vision technology, acquire multi-angle video data of a rehabilitation person using a lower limb hip-knee rehabilitation device;

[0059] S104, Based on each angle, extract key frames from the video data to form a rehabilitation process image dataset. Each angle corresponds to an image dataset. Locate the target object of the rehabilitation device in the image dataset, analyze the movement characteristics of the target object, and combine with the rehabilitation plan to divide multiple movement cycles in the usage process time period;

[0060] S106, Within one movement cycle, perform moving object detection, object segmentation, and feature extraction based on the HOG algorithm on the image datasets of multiple angles to form movement feature data for each image dataset. Taking one angle as one scale, perform multi-scale fusion on the movement feature data through a hierarchical multi-scale Transformer model to generate fused movement feature data;

[0061] S108, Based on the rehabilitation plan and the movement cycle, set the movement parameters of the moving object at multiple angles in the ideal state. Through the GAN generation model, combine the movement parameters to generate simulated features of the moving object to obtain standard movement features at multiple angles. Perform multi-scale fusion on the standard movement features through a hierarchical multi-scale Transformer model to generate standard fused feature data;

[0062] S110, Take the fused movement feature data and the standard fused feature data as two samples, calculate the similarity between the samples based on the Mahalanobis distance, and evaluate the operating state of the rehabilitation device based on the similarity to generate a rehabilitation movement index;

[0063] S112, Calculate the corresponding rehabilitation movement index for each movement cycle, and compare the rehabilitation movement index with the preset range to optimize the settings of each module of the corresponding rehabilitation device.

[0064] According to the embodiments of the present invention, the acquisition of multi-angle video data of a rehabilitation person when using a lower limb hip-knee rehabilitation device through machine vision technology is specifically as follows:

[0065] Through machine vision technology, set multi-angle video acquisition devices and apply them within the preset movement range of the rehabilitation person;

[0066] Based on the multi-angle video acquisition devices, obtain the corresponding multi-angle video data within the usage time period of a lower limb hip-knee rehabilitation device.

[0067] It should be noted that there are multiple video acquisition devices to collect multi-angle video data. The multi-angle video data includes multiple video data corresponding to multiple angles.

[0068] According to an embodiment of the present invention, for each angle, key frames are extracted from the video data to form a rehabilitation process image dataset. Each angle corresponds to an image dataset. The target object of the rehabilitation device is located in the image dataset, and the movement characteristics of the target object are analyzed. Combining with the rehabilitation plan, multiple motion cycles are divided in the usage process time period, specifically as follows:

[0069] Based on each angle, key frames are extracted from the video data to form a rehabilitation process image dataset;

[0070] The image dataset is preprocessed by image denoising, enhancement, and normalization;

[0071] Based on the rehabilitation device, target objects are set for multiple device joint positions, and the key image features of the target objects are stored;

[0072] Select an image dataset of one angle. Through the recognition module, multiple frames of images in the image dataset are recognized. The recognition process is to detect and locate the target object by combining the key image features, and analyze the movement characteristics of the target object in multiple frames of images to generate the movement data of the target object;

[0073] Analyze the movement data, combine with the rehabilitation plan to analyze the coherence of the actions of the rehabilitation personnel, and divide multiple motion cycles in the usage process time period.

[0074] It should be noted that the image dataset includes image frames in the entire time period. The recognition module can perform target detection and recognition through the YOLO recognition model. The movement data includes information such as the movement amplitude, movement speed, and movement angle of the target object at the image level. Each motion cycle corresponds to a coherent action time period, and different motion cycles correspond to the operation time periods of different rehabilitation device modules. Dividing multiple motion cycles helps to perform multi-dimensional motion state analysis on the rehabilitation device. The rehabilitation device refers to a lower limb hip-knee rehabilitation device, which generally includes multiple joint modules, and different joint modules correspond to independent control parameters, such as assist parameters, electrical stimulation parameters, and auxiliary function parameters, so as to set the rehabilitation operation state of the rehabilitation device. The target object is the multiple device joint parts.

[0075] According to an embodiment of the present invention, within one motion cycle, motion target detection, target segmentation, and feature extraction based on the HOG algorithm are performed on the image datasets of multiple angles to form motion feature data for each image dataset. Taking one angle as one scale, multi-scale fusion of the motion feature data is performed through a hierarchical multi-scale Transformer model to generate fused motion feature data, specifically as follows:

[0076] The image dataset is grayscale processed. Based on the HOG algorithm, the gradient magnitude and direction of local pixel points are calculated for each frame of the image, and multiple object contour features are extracted. Target detection and segmentation are performed from the multiple object contour features, and the target feature vector of each frame of the image is constructed;

[0077] Feature change analysis is performed on the target feature vectors of two adjacent frames of images to generate feature difference information. The target feature vector of each frame of the image and the feature difference information are integrated to form motion feature data;

[0078] For each image dataset, corresponding motion feature data is generated;

[0079] Through a hierarchical multi-scale Transformer model, based on each angle as a scale, multiple Transformer sub-modules are constructed. The multiple motion feature data are respectively input into the Transformer sub-modules for feature transformation to form multi-scale feature data;

[0080] Through the attention mechanism, the multi-scale feature data are fused and represented to generate fused motion feature data.

[0081] It should be noted that the moving target is generally the overall target when a rehabilitation person wears or uses a rehabilitation device, and is used to analyze the motion states of the rehabilitation device and the rehabilitation person during the combined use of the rehabilitation device and the rehabilitation person. The feature change analysis can be performed based on vector differences.

[0082] According to the embodiments of the present invention, each motion cycle includes corresponding fused motion feature data.

[0083] It should be noted that the fused motion feature data can effectively describe the feature information of the rehabilitation motion within a motion cycle, and this fused feature is based on multi-angle fusion analysis, which can more comprehensively extract the motion state and motion conditions during the rehabilitation motion. Further, it can compare the rehabilitation conditions in the subsequent stage and screen out the rehabilitation device modules to be optimized, so as to realize the operation optimization and personalized setting of the device to meet different rehabilitation plans and rehabilitation populations.

[0084] According to the embodiments of the present invention, based on the rehabilitation plan and the motion cycle, the motion parameters of the motion target from multiple angles are set in the ideal state. Through the GAN generation model, combined with the motion parameters, the simulated features of the motion target are generated to obtain the standard motion features from multiple angles. The standard motion features are multi-scale fused through the hierarchical multi-scale Transformer model to generate standard fused feature data, specifically:

[0085] Based on the division of the rehabilitation plan and the motion cycle, the motion parameters of the motion target from multiple angles are set in the ideal state;

[0086] According to the motion parameters, obtain multiple angular image frames of the moving target in the initial state and the motion state from the system database. For each angular image frame, it is marked as the initial image set;

[0087] Perform feature extraction on the initial image set based on the HOG algorithm to form the initial motion feature data;

[0088] Construct a GAN-based generation model, import the initial motion feature data and motion parameters into the generation model for training and learning, and cyclically generate simulated motion feature data. In each training process, analyze the object motion state based on the simulated motion feature data and the initial motion feature data, use the consistency between the object motion state and the motion parameters as the judgment of the accuracy of the generated data, and perform cyclic training until the generation model reaches the preset number of training times;

[0089] Generate standard motion features through the trained generation model;

[0090] Generate standard motion features from multiple angles based on the initial image set from multiple angles;

[0091] Perform multi-scale fusion on the standard motion features through a hierarchical multi-scale Transformer model to generate standard fusion feature data.

[0092] It should be noted that there are multiple initial image sets, and each initial image set corresponds to one angle. The initial image set is a set of multiple simple frames of the target object in the preset motion state, obtained from the system database. Subsequently, through the GAN generation model, the initial image set is used to simulate the features of the entire motion process to obtain simulated feature data from multiple angles. This simulated feature data is simulated data of coherent actions and is used for real-time feature comparison, so as to be able to evaluate the rehabilitation and motion conditions during the operation of the device. The process of multi-scale fusion of the standard motion features through the hierarchical multi-scale Transformer model is consistent with the analysis and fusion process of the fused motion feature data.

[0093] According to the embodiments of the present invention, taking the fused motion feature data and the standard fusion feature data as two samples, calculating the similarity between the samples based on the Mahalanobis distance, and evaluating the operating state of the rehabilitation device based on the similarity, generating a rehabilitation motion index, specifically:

[0094] Taking the fused motion feature data and the standard fusion feature data as two samples, and calculating the distance value between the samples based on the Mahalanobis distance;

[0095] Analyze the similarity between the samples through the distance value, and evaluate the operating state of the rehabilitation device based on the similarity, generating a rehabilitation motion index;

[0096] The rehabilitation exercise index is proportional to the similarity.

[0097] According to an embodiment of the present invention, for each exercise cycle, the corresponding rehabilitation exercise index is calculated, and by comparing the rehabilitation exercise index with a preset range, each module of the corresponding rehabilitation device is optimized and set. Specifically:

[0098] For each exercise cycle, calculation and analysis are performed to obtain the corresponding rehabilitation exercise index;

[0099] Judge whether the rehabilitation exercise index of each exercise cycle meets the preset range. If not, mark the corresponding exercise cycle as an optimized exercise cycle;

[0100] Mark the device modules operated by the rehabilitation device during the optimized exercise cycle, optimize and set the parameters of the device modules, and generate an optimization plan.

[0101] It should be noted that the optimization plan includes angle setting, assist setting, auxiliary function setting, electrical stimulation parameter setting, etc. of each module of the rehabilitation device. The device module is the joint module.

[0102] Here, it is worth mentioning that when a rehabilitation person uses a rehabilitation device, the lower limb movement state and the device operation state are a series of complex processes. Traditional technologies are difficult to effectively analyze the rehabilitation exercise process and perform multi-dimensional exercise evaluations, resulting in difficulties in achieving precise device parameter optimization and device regulation, making the rehabilitation device stay at the popular parameter settings, and it is difficult to further improve the operation efficiency and use effect of the rehabilitation device.

[0103] In the present invention, the integration of multi-scale feature analysis helps to accurately analyze the coherence, standardness, and rehabilitation state of actions under each overall exercise cycle or a movement pattern, and based on the device movement conditions from multiple angles for the rehabilitation exercise process, integrated analysis is performed, and the corresponding rehabilitation exercise index is generated to achieve an effective evaluation of the rehabilitation exercise; further, the rehabilitation exercise indices of each exercise cycle are compared and analyzed to optimize the parameters and perform feedback regulation on the corresponding device modules.

[0104] According to an embodiment of the present invention, taking the integrated motion feature data and the standard integrated feature data as two samples, and calculating the similarity between the samples based on the Mahalanobis distance, further includes:

[0105] Select two consecutive exercise cycles;

[0106] Take the integrated motion feature data corresponding to the two exercise cycles as the first sample data;

[0107] Take the standard integrated feature data corresponding to the two exercise cycles as the second sample data;

[0108] Using the DBSCAN clustering algorithm, perform density clustering analysis on the first sample data and the second sample data respectively, and form two clustering results;

[0109] The first clustering result includes N1 characteristic clustering groups;

[0110] The second clustering result includes N2 characteristic clustering groups;

[0111] If N1 is equal to N2, it is determined that the first sample data and the second sample data are similar;

[0112] Sort the N1 characteristic clustering groups and the N2 characteristic clustering groups based on the data volume of the clustering groups, and select two characteristic clustering groups from the N1 characteristic clustering groups and the N2 characteristic clustering groups in sequence to calculate the similarity, and calculate the average similarity based on the quantity N1;

[0113] The similarity calculation of the characteristic clustering groups is specifically as follows: vectorize the features of the two groups of characteristic clustering groups and evaluate and calculate the similarity of the two groups of data based on the standard Euclidean distance;

[0114] Take the average similarity as an evaluation of the operating state of the rehabilitation device, and generate a rehabilitation exercise index based on two motion cycles.

[0115] It should be noted that for the case where the rehabilitation cycle is long and the motion characteristics are complex, optionally, the similarity analysis of the fusion motion feature data and the standard fusion feature data can be performed in a clustering form. In the present invention, by integrating and analyzing the fusion motion feature data of two consecutive motion cycles, sufficient motion feature data can be integrated and used as sample data for clustering analysis. The clustering process is actually a process of classifying motions. If the classification results (the number of classifications) are the same, it can be considered that the similarity degree of the two types of feature data is relatively high, and the similarity is calculated accordingly. The Mahalanobis distance calculation is applicable to the case where there are fewer motion characteristics in each motion cycle, and the calculation efficiency is relatively high.

[0116] Figure 2 Fig. shows a block diagram of an operation optimization system for a lower limb hip-knee rehabilitation device based on machine vision according to the present invention.

[0117] The second aspect of the present invention also provides an operation optimization system 2 for a lower limb hip-knee rehabilitation device based on machine vision. The system includes: a memory 21 and a processor 22. The memory 21 includes an operation optimization program for a lower limb hip-knee rehabilitation device based on machine vision. When the operation optimization program for a lower limb hip-knee rehabilitation device based on machine vision is executed by the processor 22, the following steps are implemented:

[0118] Through machine vision technology, obtain multi-angle video data of a rehabilitation person when using a lower limb hip-knee rehabilitation device;

[0119] Based on each angle, key frames are extracted from the video data to form a rehabilitation process image data set. Each angle corresponds to an image data set. The target object of the rehabilitation device is located in the image data set, and the movement characteristics of the target object are analyzed. Combining with the rehabilitation plan, multiple motion cycles are divided in the usage process time period;

[0120] Within one motion cycle, motion target detection, target segmentation and feature extraction based on the HOG algorithm are performed on the image data sets of multiple angles to form motion feature data for each image data set. Taking one angle as one scale, the motion feature data is subjected to multi-scale fusion through a hierarchical multi-scale Transformer model to generate fused motion feature data;

[0121] Based on the rehabilitation plan and the motion cycle, the motion parameters of the motion target from multiple angles are set in the ideal state. Through the GAN generation model, combined with the motion parameters, the simulated features of the motion target are generated to obtain the standard motion features from multiple angles. The standard motion features are subjected to multi-scale fusion through a hierarchical multi-scale Transformer model to generate standard fused feature data;

[0122] Taking the fused motion feature data and the standard fused feature data as two samples, the similarity between the samples is calculated based on the Mahalanobis distance, and the operating state of the rehabilitation device is evaluated based on the similarity to generate a rehabilitation motion index;

[0123] For each motion cycle, the corresponding rehabilitation motion index is calculated, and by comparing the rehabilitation motion index with the preset range, the various modules of the corresponding rehabilitation device are optimized and set.

[0124] According to the embodiment of the present invention, the multi-angle video data of the rehabilitation personnel when using the lower limb hip-knee rehabilitation device is obtained through machine vision technology, specifically:

[0125] Through machine vision technology, video acquisition devices corresponding to multiple angles are set and applied within the preset motion range of the rehabilitation personnel;

[0126] Based on the multi-angle video acquisition devices, within the usage time period of a lower limb hip-knee rehabilitation device, the corresponding multi-angle video data is obtained.

[0127] It should be noted that there are multiple video acquisition devices to collect multi-angle video data. The multi-angle video data includes multiple video data corresponding to multiple angles.

[0128] According to an embodiment of the present invention, for each angle, key frames are extracted from the video data to form a rehabilitation process image dataset. Each angle corresponds to an image dataset. The target object of the rehabilitation device is located in the image dataset, and the movement characteristics of the target object are analyzed. In combination with the rehabilitation plan, multiple motion cycles are divided in the usage process time period, specifically as follows:

[0129] Based on each angle, key frames are extracted from the video data to form a rehabilitation process image dataset;

[0130] The image dataset is preprocessed by image denoising, enhancement, and normalization;

[0131] Based on the rehabilitation device, target objects are set for multiple device joint positions, and the key image features of the target objects are stored;

[0132] Select an image dataset of one angle. Through the recognition module, multiple frames of images in the image dataset are recognized. The recognition process is to detect and locate the target object by combining the key image features, and analyze the movement characteristics of the target object in multiple frames of images to generate the movement data of the target object;

[0133] Analyze the movement data, and in combination with the rehabilitation plan, analyze the coherence of the actions of the rehabilitation personnel, and divide multiple motion cycles in the usage process time period.

[0134] It should be noted that the image dataset includes image frames in the entire time period. The recognition module can perform target detection and recognition through the YOLO recognition model. The movement data includes information such as the movement amplitude, movement speed, and movement angle of the target object at the image level. Each motion cycle corresponds to a coherent action time period, and different motion cycles correspond to the operation time periods of different rehabilitation device modules. Dividing multiple motion cycles helps to perform multi-dimensional motion state analysis on the rehabilitation device. The rehabilitation device refers to a lower limb hip-knee rehabilitation device, which generally includes multiple joint modules, and different joint modules correspond to independent control parameters, such as assist parameters, electrical stimulation parameters, and auxiliary function parameters, which are used to set the rehabilitation operation state of the rehabilitation device. The target object is the multiple device joint parts.

[0135] According to an embodiment of the present invention, within one motion cycle, motion target detection, target segmentation, and feature extraction based on the HOG algorithm are performed on the image datasets of multiple angles to form motion feature data for each image dataset. Taking one angle as one scale, multi-scale fusion of the motion feature data is performed through a hierarchical multi-scale Transformer model to generate fused motion feature data, specifically as follows:

[0136] The image dataset is grayscale processed. Based on the HOG algorithm, the gradient magnitude and direction of local pixels of each frame of the image are calculated, and multiple object contour features are extracted. Target detection and segmentation are performed from the multiple object contour features, and the target feature vector of each frame of the image is constructed;

[0137] Feature change analysis is performed on the target feature vectors of two adjacent frames of images to generate feature difference information. The target feature vector of each frame of the image and the feature difference information are integrated to form motion feature data;

[0138] For each image dataset, the corresponding motion feature data is generated;

[0139] Through a hierarchical multi-scale Transformer model, based on each angle as a scale, multiple Transformer sub-modules are constructed. The multiple motion feature data are respectively input into the Transformer sub-modules for feature transformation to form multi-scale feature data;

[0140] Through the attention mechanism, the multi-scale feature data are fused and represented to generate fused motion feature data.

[0141] It should be noted that the moving target is generally the overall target when the rehabilitation personnel wear or use the rehabilitation equipment, and is used to analyze the motion states of the rehabilitation equipment and the rehabilitation personnel during the combined use of the rehabilitation equipment and the rehabilitation personnel. The feature change analysis can be performed based on vector differences.

[0142] According to the embodiments of the present invention, each motion cycle includes the corresponding fused motion feature data.

[0143] It should be noted that the fused motion feature data can effectively describe the feature information of the rehabilitation motion within a motion cycle, and this fused feature is fused and analyzed from multiple angles, which can more comprehensively extract the motion state and motion situation during the rehabilitation motion. Further, it can compare the rehabilitation situation in the subsequent process and screen out the rehabilitation equipment modules to be optimized, so as to realize the operation optimization and personalized setting of the equipment to meet different rehabilitation plans and rehabilitation populations.

[0144] According to the embodiments of the present invention, based on the rehabilitation plan and the motion cycle, the motion parameters of the motion target at multiple angles in the ideal state are set. Through the GAN generation model, combined with the motion parameters, the simulated features of the motion target are generated to obtain the standard motion features at multiple angles. The standard motion features are multi-scale fused through the hierarchical multi-scale Transformer model to generate the standard fused feature data, specifically:

[0145] Based on the division of the rehabilitation plan and the motion cycle, the motion parameters of the motion target at multiple angles in the ideal state are set;

[0146] According to the motion parameters, obtain multiple angular image frames of the moving target in the initial state and the motion state from the system database. For each angular image frame, label it as the initial image set;

[0147] Perform feature extraction on the initial image set based on the HOG algorithm to form initial motion feature data;

[0148] Construct a GAN-based generation model, import the initial motion feature data and motion parameters into the generation model for training and learning, cyclically generate simulated motion feature data. In each training process, analyze the object motion state based on the simulated motion feature data and the initial motion feature data, use the consistency between the object motion state and the motion parameters as the judgment of the accuracy of the generated data, and perform cyclic training until the generation model reaches the preset number of training times;

[0149] Generate standard motion features through the trained generation model;

[0150] Generate standard motion features from the initial image set at multiple angles;

[0151] Perform multi-scale fusion on the standard motion features through a hierarchical multi-scale Transformer model to generate standard fusion feature data.

[0152] It should be noted that there are multiple initial image sets, and each initial image set corresponds to one angle. The initial image set is a set of multiple simple frames of the target object in the preset motion state, obtained from the system database. Subsequently, through the GAN generation model, the initial image set is used to simulate the features of the entire motion process to obtain simulated feature data at multiple angles. This simulated feature data is simulated data of coherent actions and is used for real-time feature comparison, so as to be able to evaluate the rehabilitation and motion conditions during the operation of the device. The process of multi-scale fusion of the standard motion features through the hierarchical multi-scale Transformer model is consistent with the analysis and fusion process of the fused motion feature data.

[0153] According to the embodiments of the present invention, use the fused motion feature data and the standard fusion feature data as two samples, calculate the similarity between the samples based on the Mahalanobis distance, and evaluate the operating state of the rehabilitation device based on the similarity to generate a rehabilitation motion index. Specifically:

[0154] Use the fused motion feature data and the standard fusion feature data as two samples, and calculate the distance value between the samples based on the Mahalanobis distance;

[0155] Analyze the similarity between the samples through the distance value, and evaluate the operating state of the rehabilitation device based on the similarity to generate a rehabilitation motion index;

[0156] The rehabilitation exercise index is directly proportional to the similarity.

[0157] According to an embodiment of the present invention, for each exercise cycle, a corresponding rehabilitation exercise index is calculated, and by comparing the rehabilitation exercise index with a preset range, each module of the corresponding rehabilitation device is optimized and set. Specifically:

[0158] For each exercise cycle, calculation and analysis are performed to obtain a corresponding rehabilitation exercise index;

[0159] It is determined whether the rehabilitation exercise index of each exercise cycle meets the preset range. If not, the corresponding exercise cycle is marked as an optimized exercise cycle;

[0160] Mark the device modules operated by the rehabilitation device during the optimized exercise cycle, optimize and set the parameters of the device modules, and generate an optimization plan.

[0161] It should be noted that the optimization plan includes angle setting, assistance setting, auxiliary function setting, electrical stimulation parameter setting, etc. for each module of the rehabilitation device. The device module is the joint module.

[0162] Here, it is worth mentioning that when a rehabilitation person uses a rehabilitation device, the lower limb movement state and the device operation state are a series of complex processes. Traditional technologies are difficult to effectively analyze the rehabilitation exercise process and conduct multi-dimensional motion assessments, resulting in difficulties in achieving precise device parameter optimization and device regulation, making the rehabilitation device stay at a popular parameter setting, and it is difficult to further improve the operation efficiency and use effect of the rehabilitation device.

[0163] In the present invention, integrating multi-scale feature analysis helps to accurately analyze the coherence, standardness, and rehabilitation state of actions under each overall exercise cycle or a motion mode, and based on the device motion conditions from multiple angles for the rehabilitation exercise process, integrated analysis is carried out, and a corresponding rehabilitation exercise index is generated to realize the effectiveness assessment of the rehabilitation exercise; further, the rehabilitation exercise indexes of each exercise cycle are compared and analyzed to optimize the parameters and perform feedback regulation on the corresponding device modules.

[0164] The third aspect of the present invention also provides a computer-readable storage medium, which includes an operation optimization program for a lower limb hip-knee rehabilitation device based on machine vision. When the operation optimization program for the lower limb hip-knee rehabilitation device based on machine vision is executed by a processor, the steps of the operation optimization method for the lower limb hip-knee rehabilitation device based on machine vision as described in any one of the above are realized.

[0165] The present invention discloses an operation optimization method for a lower limb hip-knee rehabilitation device based on machine vision. First, multi-angle video data of a rehabilitation person using the device is obtained, key frames are extracted and the target object is located, and the movement characteristics are analyzed and the movement cycle is divided; the HOG algorithm is used to detect, segment and extract movement characteristics, and multi-scale fusion is performed through a Transformer model to generate fused movement characteristic data; based on the rehabilitation plan, a GAN generation model is used to simulate the standard movement characteristics in an ideal state, and multi-scale fusion is performed to generate standard fused characteristic data; the similarity between the fused movement characteristic data and the standard fused characteristic data is calculated, the operation state of the rehabilitation device is evaluated and a rehabilitation movement index is generated; based on the rehabilitation movement index, the parameters of the device module are optimized. Through the present invention, the coherence, standardization and corresponding rehabilitation state of the operation of the rehabilitation device can be accurately analyzed, the rehabilitation effect can be effectively evaluated, and the dynamic optimization and real-time adjustment of the device parameters can be realized.

[0166] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined, or can be integrated into another system, or some features can be ignored, or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed with each other can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical, or other forms.

[0167] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units; they can be located in one place or distributed to multiple network units; some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0168] In addition, each functional unit in the embodiments of the present invention can be all integrated in a processing unit, or each unit can be separately used as a unit, or two or more units can be integrated in a unit; the above integrated units can be implemented in the form of hardware, or in the form of a combination of hardware and software functional units.

[0169] Those of ordinary skill in the art can understand that all or part of the steps to implement the above method embodiments can be completed by hardware related to program instructions. The foregoing program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps including those of the above method embodiments. The foregoing storage medium includes: various media that can store program codes, such as removable storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs.

[0170] Alternatively, if the above integrated unit of the present invention is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the embodiments of the present invention, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in the various embodiments of the present invention. The foregoing storage medium includes: various media that can store program codes, such as removable storage devices, ROM, RAM, magnetic disks, or optical discs.

[0171] As described above, the above are only the specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention.

Claims

1. An operation optimization method for a lower limb hip-knee rehabilitation device based on machine vision, characterized in that Including: By means of machine vision technology, multi-angle video data of a rehabilitation person using a lower limb hip-knee rehabilitation device is obtained; Based on each angle, key frames are extracted from the video data to form a rehabilitation process image data set. Each angle corresponds to an image data set. The target object of the rehabilitation device is located in the image data set, and the movement characteristics of the target object are analyzed. Combining with the rehabilitation plan, multiple motion cycles are divided in the time period of the use process; Within one motion cycle, motion target detection, target segmentation and feature extraction based on the HOG algorithm are performed on the image data sets of multiple angles to form motion feature data for each image data set. Taking one angle as one scale, the motion feature data is subjected to multi-scale fusion through a hierarchical multi-scale Transformer model to generate fused motion feature data. Specifically, the image data set is grayscale processed, the gradient magnitude and direction of local pixel points of each frame of image are calculated based on the HOG algorithm, and multiple object contour features are extracted. Target detection and segmentation are performed from multiple object contour features to construct the target feature vector of each frame of image; Feature change analysis is performed on the target feature vectors of two adjacent frames of images to generate feature difference information, and the target feature vector of each frame of image and the feature difference information are integrated to form motion feature data; For each image data set, corresponding motion feature data is generated; Through a hierarchical multi-scale Transformer model, based on each angle as one scale, multiple Transformer sub-modules are constructed, and multiple motion feature data are respectively input into the Transformer sub-modules for feature transformation to form multi-scale feature data; Through the attention mechanism, the multi-scale feature data is subjected to feature fusion representation to generate fused motion feature data; Based on the rehabilitation plan and the motion cycle, the motion parameters of the motion target in the ideal state are set for multiple angles. Through the GAN generation model, combined with the motion parameters, the simulated features of the motion target are generated to obtain the standard motion features of multiple angles. The standard motion features are subjected to multi-scale fusion through a hierarchical multi-scale Transformer model to generate standard fused feature data; The fused motion feature data and the standard fused feature data are used as two samples, and the similarity between the samples is calculated based on the Mahalanobis distance, and the operating state of the rehabilitation device is evaluated based on the similarity to generate a rehabilitation motion index; The corresponding rehabilitation motion index is calculated for each motion cycle, and the corresponding rehabilitation device modules are optimized by comparing the rehabilitation motion index with the preset range.

2. The operation optimization method of a lower limb hip and knee rehabilitation device based on machine vision according to claim 1, wherein The obtaining of the multi-angle video data of the rehabilitation person using the lower limb hip-knee rehabilitation device by means of machine vision technology is specifically as follows: By means of machine vision technology, a multi-angle video acquisition device is set and applied within the preset motion range of the rehabilitation person; Based on the multi-angle video acquisition device, the corresponding multi-angle video data is obtained within the use time period of a lower limb hip-knee rehabilitation device.

3. A method for optimizing the operation of a lower limb hip and knee rehabilitation device based on machine vision according to claim 1, characterized in that, Based on each angle, key frames are extracted from the video data to form a rehabilitation process image dataset. Each angle corresponds to an image dataset. The target object of the rehabilitation device is located in the image dataset, and the movement characteristics of the target object are analyzed. Combining with the rehabilitation plan, multiple motion cycles are divided in the usage process time period, specifically as follows: Based on each angle, key frames are extracted from the video data to form a rehabilitation process image dataset; The image dataset is preprocessed by image denoising, enhancement, and standardization; Based on the rehabilitation device, target objects are set for multiple device joint positions, and the key image features of the target objects are stored; Select an image dataset of one angle. Through the recognition module, multiple frames of images in the image dataset are recognized. The recognition process is to detect and locate the target object by combining the key image features, and analyze the movement characteristics of the target object in multiple frames of images to generate the movement data of the target object; Analyze the movement data, combine with the rehabilitation plan to analyze the coherence of the actions of the rehabilitation personnel, and divide multiple motion cycles in the usage process time period.

4. A method for optimizing the operation of a lower limb hip-knee rehabilitation device based on machine vision according to claim 1, characterized in that Each motion cycle includes corresponding fused motion feature data.

5. A method for optimizing the operation of a lower limb hip-knee rehabilitation device based on machine vision according to claim 1, characterized in that, Based on the rehabilitation plan and the motion cycle, set the motion parameters of the motion target from multiple angles in the ideal state. Through the GAN generation model, combine the motion parameters to generate the simulated feature of the motion target, and obtain the standard motion features from multiple angles. Through the hierarchical multi-scale Transformer model, multi-scale fusion of the standard motion features is performed to generate the standard fusion feature data, specifically as follows: Based on the rehabilitation plan and the motion cycle division situation, set the motion parameters of the motion target from multiple angles in the ideal state; According to the motion parameters, multiple angle image frames of the motion target in the initial state and the motion state are obtained from the system database. For each angle image frame, it is marked as the initial image set; The initial image set is subjected to feature extraction based on the HOG algorithm to form the initial motion feature data; Construct a GAN-based generation model, import the initial motion feature data and the motion parameters into the generation model for training and learning, and cyclically generate simulated motion feature data. In each training process, analyze the object motion state according to the simulated motion feature data and the initial motion feature data, and use the consistency between the object motion state and the motion parameters as the judgment of the accuracy of the generated data, and perform cyclic training until the generation model reaches the preset number of training times; Generate standard motion features through the trained generation model; Generate standard motion features from multiple angles based on the initial image set from multiple angles; Through the hierarchical multi-scale Transformer model, multi-scale fusion of the standard motion features is performed to generate the standard fusion feature data.

6. The operation optimization method of a lower limb hip and knee rehabilitation device based on machine vision according to claim 1, characterized in that Take the fused motion feature data and the standard fusion feature data as two samples, calculate the similarity between the samples based on the Mahalanobis distance, and evaluate the operating state of the rehabilitation device based on the similarity to generate a rehabilitation motion index, specifically as follows: Take the fused motion feature data and the standard fusion feature data as two samples, and calculate the distance value between the samples based on the Mahalanobis distance; Analyze the similarity between samples through distance values, and evaluate the operating status of the rehabilitation device based on the similarity to generate a rehabilitation exercise index; The rehabilitation exercise index is proportional to the similarity.

7. A method for optimizing the operation of a lower limb hip and knee rehabilitation device based on machine vision according to claim 1, characterized in that, For each exercise cycle, calculate the corresponding rehabilitation exercise index, and compare the rehabilitation exercise index with a preset range to optimize the settings of each module of the corresponding rehabilitation device. Specifically: Perform calculation and analysis for each exercise cycle to obtain the corresponding rehabilitation exercise index; Determine whether the rehabilitation exercise index of each exercise cycle meets the preset range. If not, mark the corresponding exercise cycle as an optimized exercise cycle; Mark the device modules operated by the rehabilitation device during the optimized exercise cycle, optimize the parameter settings of the device modules, and generate an optimization plan.

8. An operation optimization system for a lower limb hip and knee rehabilitation device based on machine vision, characterized in that, The system includes: a memory and a processor. The memory includes an optimization program for the operation of a lower limb hip-knee rehabilitation device based on machine vision. When the optimization program for the operation of the lower limb hip-knee rehabilitation device based on machine vision is executed by the processor, the following steps are implemented: Through machine vision technology, obtain multi-angle video data of a rehabilitation person when using a lower limb hip-knee rehabilitation device; Based on each angle, extract key frames from the video data to form a rehabilitation process image data set. Each angle corresponds to an image data set. Locate the target object of the rehabilitation device in the image data set, analyze the movement characteristics of the target object, and combine the rehabilitation plan to divide the usage process time period into multiple exercise cycles; Within an exercise cycle, perform motion target detection, target segmentation, and feature extraction on the image data sets of multiple angles based on the HOG algorithm to form motion feature data for each image data set. Using one angle as one scale, perform multi-scale fusion on the motion feature data through a hierarchical multi-scale Transformer model to generate fused motion feature data. Specifically, perform grayscale processing on the image data set, calculate the gradient magnitude and direction of local pixel points for each frame of image based on the HOG algorithm, extract multiple object contour features, perform target detection and segmentation from the multiple object contour features, and construct the target feature vector of each frame of image; Perform feature change analysis on the target feature vectors of two adjacent frames of images to generate feature difference information, and integrate the target feature vector of each frame of image with the feature difference information to form motion feature data; Generate corresponding motion feature data for each image data set; Through a hierarchical multi-scale Transformer model, based on each angle as one scale, construct multiple Transformer sub-modules, and input the multiple motion feature data into the Transformer sub-modules respectively for feature transformation to form multi-scale feature data; Through the attention mechanism, perform feature fusion representation on the multi-scale feature data to generate fused motion feature data; Based on the rehabilitation plan and the exercise cycle, set the exercise parameters from multiple angles in the ideal state of the exercise goal. Through the GAN generation model, combine the exercise parameters to generate the simulated features of the exercise goal, and obtain the standard exercise features from multiple angles. Through the hierarchical multi-scale Transformer model, perform multi-scale fusion on the standard exercise features to generate the standard fusion feature data; Take the fusion exercise feature data and the standard fusion feature data as two samples, calculate the similarity between the samples based on the Mahalanobis distance, and evaluate the operating state of the rehabilitation device based on the similarity to generate the rehabilitation exercise index; Calculate the corresponding rehabilitation exercise index for each exercise cycle, and compare the rehabilitation exercise index with the preset range to optimize the settings of each module of the corresponding rehabilitation device.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes an optimization program for the operation of a lower limb hip-knee rehabilitation device based on machine vision. When the optimization program for the operation of a lower limb hip-knee rehabilitation device based on machine vision is executed by a processor, the steps of the optimization method for the operation of a lower limb hip-knee rehabilitation device based on machine vision according to any one of claims 1 to 7 are implemented.

Images