A gait recognition method based on brain cortex algorithm

Abstract

The application discloses a gait recognition method based on a brain cortex algorithm, and comprises the following steps: acquiring a gait image; performing gait segmentation on the gait image to obtain a binary contour image; performing gait image coding on the binary contour image to obtain an encoding result of the gait segmentation binary image; performing gait image sparse discrete representation on the encoding result of the gait segmentation binary image to obtain a gait segmentation image sparse representation; and comparing the gait segmentation image sparse representation with a pre-stored gait sparse discrete representation to obtain a gait recognition result. In the gait segmentation process, time information is aggregated, the stability of foreground extraction is enhanced, the extraction speed is improved, and the gait image can be effectively extracted in a complex environment. The binary contour image is subjected to first and second spatial pooling, so that the reliability of recognition is improved, the calculation is reduced, and the calculation efficiency is improved. The application can be widely applied to the technical field of image recognition.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image recognition, and in particular to a gait recognition method based on a cerebral cortex algorithm. BACKGROUND

[0002] When the number of people is huge, the existing technical means for identifying the identity of the person becomes very difficult, the workload is large, the investigation efficiency is low, the work is tedious, and a lot of manpower and material resources are needed. In the face of such a large number of personal identity recognition problems, we urgently need to propose related technical means to assist in investigation and improve the work quality and efficiency of investigation and identification.

[0003] Gait recognition technology uses walking posture for personal identity recognition. Compared with fingerprint, face, iris and other biometric recognition technologies, gait recognition technology has the advantages of non-contact, non-invasion, easy perception, difficult to hide and disguise, etc. In the case of long distance, gait recognition is the most feasible technology at present. As a world-leading biometric technology, gait recognition technology is currently still in the experimental stage, and many technical problems need to be broken through, but the research on this technology still has a significant meaning that cannot be ignored.

[0004] Gait recognition is mainly divided into two categories: model-based gait recognition and appearance-based gait recognition.

[0005] The model-based method needs to model the human body. The existing related technology needs to respectively perform linear modeling, 2D image modeling and 3D modeling on the human body in a clock type, but the problem is that a high-quality data set needs to be established, and the model building process is relatively difficult, especially when the resolution of the image is low, the reliability of modeling is greatly reduced.

[0006] The appearance-based method directly tries to find gait features from gait sequence videos. This method usually performs gait segmentation on image frames, extracts the contour of the person, and directly inputs the normalized contour image into the algorithm. In addition, a well-known method is to generate a gait energy map from a series of contour images, and then integrate the time and space information before learning the algorithm. The disadvantage of the appearance-based method is that it is difficult to effectively extract the contour in a complex environment, and it is greatly affected by clothing and viewing angle. SUMMARY

[0007] Therefore, the embodiment of the present application provides a gait recognition method based on a cerebral cortex algorithm with high reliability.

[0008] In one aspect, the embodiment of the present application provides a gait recognition method based on a cerebral cortex algorithm, comprising:

[0009] obtaining a gait image;

[0010] performing gait segmentation on the gait image to obtain a binary contour image;

[0011] encoding the binary contour image to obtain an encoding result of the gait segmentation binary image;

[0012] sparse discrete representation of the gait segmentation binary image to obtain a sparse representation of the gait segmentation image;

[0013] comparing the sparse discrete representation of the gait segmentation image with a pre-stored sparse discrete representation of gait to obtain a gait recognition result.

[0014] Optionally, the step of performing gait segmentation on the gait image to obtain a binary contour image comprises:

[0015] model solidification on the RVM algorithm model;

[0016] setting input parameters of the RVM algorithm model after model solidification, the input parameters including a running device, a number of cycles, and a down-sampling ratio;

[0017] inputting the gait image into the RVM algorithm model to perform gait segmentation and obtain a binary contour image.

[0018] Optionally, the step of encoding the binary contour image to obtain an encoding result of the gait segmentation binary image comprises:

[0019] inputting the binary contour image into a first spatial pool to determine candidate neurons connected to input data;

[0020] adding weights to the candidate neurons to obtain activated neurons;

[0021] updating the activated neurons to obtain an encoding result of the gait segmentation binary image.

[0022] Optionally, the step of updating the activated neurons to obtain an encoding result of the gait segmentation binary image comprises:

[0023] updating weights of the activated neurons based on a first updating formula;

[0024] updating an enhancement coefficient of the activated neurons based on a second updating formula;

[0025] updating a constraint radius of the activated neurons based on a third updating formula.

[0026] Optionally, the step of sparse discrete representation of the encoding result of the gait segmentation binary image to obtain a sparse representation of the gait segmentation image comprises:

[0027] Adjust the neuron activation rate of the second spatial pool;

[0028] Input the encoding result of the gait segmentation binary image into the adjusted second spatial pool, and obtain a gait segmentation image sparse representation through sparse discrete representation of the gait image.

[0029] Optionally, the step of comparing the gait segmentation image sparse representation with the pre-stored gait sparse discrete representation to obtain a gait recognition result comprises:

[0030] Store the pre-stored gait sparse discrete representation in the form of a dictionary;

[0031] Exhaustively compare the gait segmentation image sparse representation with the pre-stored gait sparse discrete representation to obtain a comparison similarity;

[0032] Compare the comparison similarity with a preset comparison similarity threshold to obtain a gait recognition result; when the comparison similarity is greater than or equal to the preset comparison similarity threshold, it is determined that the target person is recognized; when the comparison similarity is less than the preset comparison similarity threshold, it is determined that the target person is not recognized.

[0033] Optionally, the step of inputting the gait image into the RVM algorithm model for gait segmentation to obtain a binary contour image comprises:

[0034] Extract features and perform semantic segmentation on the gait image through a feature encoder of the RVM algorithm model to obtain a first contour image;

[0035] Aggregate time information of the first contour image through a recurrent decoder of the RVM algorithm model to obtain a second contour image;

[0036] Perform high-resolution prediction on the second contour image through a depth-guided filter of the RVM algorithm model to obtain a binary contour image.

[0037] In another aspect, the embodiment of the present application also provides a gait recognition device based on a brain cortex algorithm, comprising:

[0038] A first module for acquiring a gait image;

[0039] A second module for performing gait segmentation on the gait image to obtain a binary contour image;

[0040] A third module for encoding the binary contour image through a gait image to obtain an encoding result of a gait segmentation binary image;

[0041] a fourth module configured to obtain a gait segmentation image sparse representation by sparse discrete representation of the gait image, through sparse discrete representation of the gait image;

[0042] a fifth module configured to obtain a gait recognition result by comparison of the gait segmentation image sparse representation with a pre-stored gait sparse discrete representation.

[0043] In another aspect, the embodiment of the present application further provides an electronic device comprising a processor and a memory;

[0044] The memory is configured to store a program.

[0045] The processor executes the program to realize the gait recognition method based on the brain cortex algorithm.

[0046] In another aspect, the embodiment of the present application further provides a computer readable storage medium, which stores a program, and the program is executed by a processor to realize the gait recognition method based on the brain cortex algorithm.

[0047] The embodiment of the present application further discloses a computer program product or a computer program, which comprises computer instructions stored in a computer readable storage medium. A processor of a computer device can read the computer instructions from the computer readable storage medium, and the processor executes the computer instructions to enable the computer device to execute the foregoing method.

[0048] The embodiment of the present application at least has the following advantages: the gait image is segmented to obtain a binary contour image, the time information is aggregated in the segmentation process, the stability of foreground extraction is enhanced, the extraction speed is accelerated, and the contour can be effectively extracted in a complex environment; the binary contour image is encoded to obtain an encoding result of the gait segmentation binary image, the activation rate of neurons is adjusted in the gait image encoding process, the amplification effect on small differences can be weakened, and the reliability of the result is improved; the gait segmentation image sparse representation is obtained through sparse discrete representation of the gait image, the calculation can be reduced, and the calculation efficiency is improved. BRIEF DESCRIPTION OF DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative effort.

[0050] Figure 1is a step diagram of a gait recognition method based on a brain cortex algorithm provided by the embodiment of the present application.

[0051] Figure 2 is a module diagram of a gait recognition device based on a brain cortex algorithm provided by the embodiment of the present application.

[0052] Figure 3 is a flowchart of a gait recognition method based on a brain cortex algorithm provided by the embodiment of the present application.

[0053] Figure 4 is a flowchart of a brain cortex learning algorithm provided by the embodiment of the present application. DETAILED DESCRIPTION

[0054] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application is further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0055] In view of the problems in the prior art, the embodiment of the present application provides a gait recognition method based on a brain cortex algorithm as shown in Figure 1 The method comprises steps 100 to 500.

[0056] Step 100: Obtain a gait image.

[0057] Specifically, first, the target person is detected by the camera device, the HOG feature (directional gradient histogram) is combined with the SVM (support vector machine) method to realize the person detection, and the gait image is obtained. The embodiment of the present application calls the trained model in opencv to detect the person, and specifically calls the detectMultiScale function in opencv HOG, which contains 8 input parameters and 2 output parameters. The two mandatory input parameters are img (input image) and foundLocations (target storage location), and the six optional input parameters are hitThreshold (threshold value of the distance of the feature to the SVM hyperplane), winStride (step length of the HOG detection window movement), padding (peripheral pixels), scale (multiple scales of the image), finalThreshold (final threshold value) and useMeanShiftGrouping (eliminate overlap), and the two output parameters are found (target position) and w (target possibility value).

[0058] The important optional parameters are winStride and scale, which greatly affect the detection accuracy and detection speed. In the embodiment of the application, only the winStride and scale parameters are configured. According to the experimental results, the actual detection speed and detection accuracy are comprehensively considered, and the parameters are configured as winStride=(4, 4) and scale=1.5. After the function input is configured, the function is called to obtain the detection target position and the possibility value. The threshold value of the possibility value is judged to exclude objects with low possibility. The possibility threshold value of the embodiment of the application is 1. After the appearance of the person is detected, the system starts to continuously record 25 frame gait sequence images, and stores the gait images in a specified folder to obtain the gait images.

[0059] Step 200: performing gait segmentation on the gait image to obtain a binary contour image.

[0060] Specifically, the step of performing gait segmentation on the gait image to obtain a binary contour image can specifically include steps 210 to 230:

[0061] Step 210: model solidification is performed on the RVM algorithm model.

[0062] Specifically, the RVM (Relevance Vector Machine) algorithm model is a Bayesian sparse kernel algorithm for regression and classification. The RVM algorithm usually regards a picture I as a linear combination of foreground F and background B, and the relationship expression is:

[0063] I = alpha F + (1-alpha) B

[0064] The background can be replaced by extracting the foreground F and the coefficient alpha, but for the gait segmentation of the application, only the foreground F needs to be extracted. The RVM algorithm extracts the image foreground based on the neural network, compared with the existing extraction method, it adds time information, enhances the stability of foreground extraction, and speeds up the extraction speed.

[0065] When the RVM algorithm is applied for gait segmentation in the embodiment of the application, a lightweight model under the TorchScript framework is adopted. The model is an officially trained model. When the model is used in the application, the downloaded model is first loaded, and then the freeze function is called to solidify the model, which can speed up the model running speed.

[0066] Step 220: setting the input parameters of the RVM algorithm model after model solidification, the input parameters including running device, cycle number and down-sampling ratio.

[0067] Specifically, after the model solidification of the RVM algorithm model is completed, the input parameters of the model are set, and the input parameters to be set include a running device (GPU or CPU), a loop number and a downsampling ratio. For example, the running device can be set as the CPU, the loop number can be set as 4, and the downsampling ratio can be set as 0.25.

[0068] Step 230: inputting the gait image into the RVM algorithm model for gait segmentation to obtain a binary contour image.

[0069] Specifically, the step of inputting the gait image into the RVM algorithm model for gait segmentation to obtain a binary contour image specifically includes steps 231 to 233.

[0070] Step 231: performing feature extraction and semantic segmentation on the gait image by a feature encoder of the RVM algorithm model to obtain a first contour image.

[0071] Specifically, the feature encoder performs feature extraction and semantic segmentation on a single frame of the gait image based on a semantic segmentation network to obtain the first contour image.

[0072] Step 232: aggregating time information of the first contour image by a loop decoder of the RVM algorithm model to obtain a second contour image.

[0073] Specifically, the loop decoder adopts a loop architecture, and the loop mechanism can learn information to be retained and information to be ignored in a continuous video stream. A multi-scale ConvGRU is used to aggregate time information, and the loop architecture is used to aggregate time information in the image, which significantly improves the time continuity and image segmentation quality.

[0074] Step 233: performing high-resolution prediction on the second contour image by a deep guided filter of the RVM algorithm model to obtain a binary contour image.

[0075] Specifically, when the RVM algorithm model processes a high-resolution image, the deep guided filter is used to complete high-resolution prediction. When processing a high-resolution image, the second contour image is inputted, down-sampled, and then the low-resolution alpha, foreground and label, and the high-resolution input frame are inputted into the deep guided filter to generate high-resolution alpha and foreground. The deep guided filter applies the idea of guided filtering to the network, first learns on a small image, then realizes up-sampling on the generated small image through the guided filter to learn the guided image of a specific task, and finally restores the large image to obtain the binary contour image.

[0076] Step 300: encoding the binary contour image through the gait image to obtain an encoding result of the gait segmentation binary image.

[0077] Specifically, the step of encoding the binary contour image through the gait image to obtain an encoding result of the gait segmentation binary image comprises steps 310 to 330.

[0078] Step 310: inputting the binary contour image into a first spatial pool to determine candidate neurons connected to the input data.

[0079] Specifically, the binary contour image obtained by gait segmentation is a two-dimensional matrix composed of 0 and 1, which is similar to the sparse discrete representation (SDR) generated by the encoder, and similar human input will also obtain similar contour output, so the application proposes to add a first spatial pool as a sparse discrete representation encoder of the gait segmentation binary image.

[0080] Specifically, the first spatial pool is a two-dimensional neural network with a total of 64x32x1=2048 neurons, one neuron in each neuron column, and the first spatial pool accepts binary contour image input and randomly determines a candidate neuron set connected to the input data according to formula (1).

[0081]

[0082] wherein x j is the jth neuron in the first spatial pool, is the ith neuron of the input image, and γ is the radius of the circle centered at I(·) is a description function, which outputs 1 if and only if x j is located in the circle centered at with a radius of γ, and the set Π is a candidate neuron set.

[0083] Step 320: adding weights to the candidate neurons to obtain activated neurons.

[0084] Specifically, the candidate neurons in the set Π are given random weights D ij When , the candidate neuron becomes a neuron connected to the input, wherein is the weight threshold of the first spatial pool, which can be 0.85, at this time, the neuron connected to the input is called an activated neuron. The number of activated neurons is less than or equal to the neuron activation rate condition, and the neuron activation rate of the first spatial pool is 7% (135). The purpose is to weaken its amplification effect on small differences; each activated neuron is multiplied by a connection strengthening coefficient bi (bi>0) to strengthen its activated state; in order to ensure the sparsity of the spatial pool output, the activated neurons will limit the neurons in the range of r in h to be in the activated state.

[0085] Step 330: updating the activated neuron to obtain an encoding result of the gait segmentation binary image.

[0086] Specifically, the activated neuron is updated to a preset update period, and an encoding result of the gait segmentation binary image is output.

[0087] Specifically, the step of updating the activated neuron to obtain an encoding result of the gait segmentation binary image specifically includes steps 331 to 333:

[0088] Step 331: updating the weight of the activated neuron based on a first update formula.

[0089] Specifically, in the neural network learning stage, the candidate neuron weight in the set Π is updated according to the first update formula (Hebbian learning algorithm), and the first update formula is as follows:

[0090]

[0091] wherein, ΔD dec is a decreasing bias of the weight of the activated neuron, ΔD inc is an increasing bias of the weight of the activated neuron, D ij is a random weight of the candidate neuron, W ij is the updated weight of the activated neuron; ΔD inc = 0.04, ΔD dec = 0.005.

[0092] Step 332: updating the reinforcement coefficient of the activated neuron based on a second update formula.

[0093] Specifically, the reinforcement coefficient of the activated neuron is updated by the second update formula, and the expression of the second update formula is:

[0094]

[0095] wherein, b i is the reinforcement coefficient of the activated neuron, β = 3 is an adjustment coefficient, T = 1000 is an update period, and a i (t) represents the activation state of the neuron i in the set Π at time t, and Ni is the total number of activated neurons at time t.

[0096] Step 333: updating the constraint radius of the activated neuron based on a third update formula.

[0097] Specifically, the reinforcement coefficient of the activated neuron is updated by the third update formula, and the expression of the third update formula is:

[0098] r inh = (CS avg x PI col - 1) / 2

[0099] where r inh is the constraint radius of the active neuron, CSavg is the number of input pixels connected to each active neuron, and PIcol is the total number of active neurons connected to all pixels of the input image.

[0100] Step 400: The encoding result of the gait segmentation binary image is represented by sparse discrete representation of gait image, to obtain sparse representation of gait segmentation image.

[0101] Specifically, the step of representing the encoding result of the gait segmentation binary image by sparse discrete representation of gait image to obtain sparse representation of gait segmentation image specifically includes steps 410 to 420:

[0102] Step 410: Adjust the activation rate of neurons of the second spatial pool.

[0103] Specifically, the encoding result of the gait segmentation binary image obtained through the first spatial pool does not have sparsity, so it needs to be represented by sparse discrete representation of gait image through the second spatial pool to obtain sparse representation of gait segmentation image. The network structure of the second spatial pool is the same as that of the first spatial pool, which is a two-dimensional neural network with a total of 64x32x1=2048 neurons, and the network parameters and learning process are similar, but the activation rate of neurons of the second spatial pool needs to be adjusted to 2% (the activation rate of neurons of the first spatial pool is 7%), to ensure that the output of the second spatial pool is sparse representation.

[0104] Step 420: Input the encoding result of the gait segmentation binary image into the adjusted second spatial pool, and represent the gait image by sparse discrete representation to obtain sparse representation of gait segmentation image.

[0105] Specifically, the encoding result of the gait segmentation binary image is input into the adjusted second spatial pool, and the gait image is represented by sparse discrete representation to obtain sparse representation of gait segmentation image.

[0106] Step 500: Compare the sparse representation of gait segmentation image with the pre-stored sparse discrete representation of gait to obtain gait recognition result.

[0107] Specifically, the step of comparing the sparse representation of gait segmentation image with the pre-stored sparse discrete representation of gait to obtain gait recognition result specifically includes steps 510 to 530:

[0108] Step 510: store the pre-stored gait sparse discrete representation in the form of a dictionary.

[0109] Specifically, the pre-stored gait sparse discrete representation is stored in the form of a dictionary.

[0110] Step 520: exhaustively compare the gait segmentation image sparse representation with the pre-stored gait sparse discrete representation to obtain a comparison similarity.

[0111] Specifically, the gait segmentation image sparse representation is exhaustively compared with the pre-stored gait sparse discrete representation, and the process is similar to the direct comparison between contours to obtain a comparison similarity of comparison, which is used to measure the similarity between the gait segmentation image sparse representation obtained by gait recognition and the pre-stored gait sparse discrete representation.

[0112] Step 530: compare the comparison similarity with a preset comparison similarity threshold to obtain a gait recognition result; when the comparison similarity is greater than or equal to the preset comparison similarity threshold, it is determined that the target person is recognized; when the comparison similarity is less than the preset comparison similarity threshold, it is determined that the target person is not recognized.

[0113] Specifically, the preset comparison similarity threshold is used to compare with the comparison similarity, and is set by itself according to actual conditions; when the comparison similarity is less than the preset comparison similarity threshold, it is determined that the gait image belongs to a stranger, and the target object is not detected; when the comparison similarity is greater than or equal to the preset comparison similarity threshold, the person identification corresponding to the gait sparse discrete representation with the highest comparison similarity is found, and the person identification can determine a unique target person, including name, age, birth date, ID number, etc. For example, the comparison similarity threshold is set to 50%, when the obtained comparison similarity is less than 50%, it is determined that the gait image belongs to a stranger, and it is determined that the target person is not recognized, when the comparison similarity is greater than or equal to 50%, the person identification corresponding to the gait sparse discrete representation with the highest comparison similarity is found, and it is determined that the target person is recognized.

[0114] The application will be described below with reference to the accompanying drawings: Figure 3 The application will be described below with reference to the accompanying drawings:

[0115] 1. First, the gait image is acquired. The detectMultiScale function in opencv HOG is called to detect the person, the position of the detection target and its possibility value are obtained, the possibility is thresholded, when the possibility is less than threshold 1, it is discarded, when the possibility is greater than or equal to 1, it is accepted, and objects with low possibility are excluded; then the gait sequence is recorded, 25 frames of gait sequence images are continuously recorded, and these gait images are stored in a specified folder;

[0116] 2. Then the gait image is segmented by the RVM algorithm model to obtain a binary contour image. First, the RVM algorithm model is solidified; then the input parameters of the RVM algorithm model after model solidification are set, the input parameters include running equipment, cycle number and downsampling ratio; finally, the gait image is input into the RVM algorithm model for gait segmentation to obtain a binary contour image;

[0117] 3. Then the binary contour image is obtained by the cerebral cortex learning algorithm to obtain the prediction result output, that is, the gait recognition result. Referring to Figure 4 , first, the binary contour image is encoded by the gait image to obtain the encoding result of the gait segmentation binary image, the gait image encoding process first initializes the candidate neuron set connected to the input, then judges whether the candidate neuron covers the entire binary contour image, if not, the candidate neuron set connected to the input is initialized again, if yes, the weight of the candidate neuron is increased, the activated neuron is determined, and the weight of the activated neuron is updated and the reinforcement coefficient and the constraint radius of the activated neuron are updated until the learning iteration number reaches the learning period, and the encoding result of the gait segmentation binary image is output;

[0118] 4. Then the encoding result of the gait segmentation binary image is represented by sparse discrete representation to obtain the sparse representation of the gait segmentation image; finally, the sparse representation of the gait segmentation image is compared with the pre-stored gait sparse discrete representation to obtain the gait recognition result.

[0119] In summary, the gait recognition method based on the cerebral cortex algorithm has the following advantages:

[0120] 1. In the embodiment of the application, the gait image is segmented to obtain a binary contour image, the time information is aggregated through the loop decoder in the segmentation process, the stability of foreground extraction is enhanced, the extraction speed is accelerated, and significant improvement is achieved in time continuity and image segmentation quality, and the contour can be effectively extracted in a complex environment;

[0121] 2、The neuron activation rate set in the first spatial pool can weaken the amplification effect on small differences, the neuron activation rate of the second spatial pool is adjusted, sparse representation of gait segmentation images is obtained, calculation can be reduced, and calculation efficiency is improved;

[0122] 3、The embodiment of the application can establish and update the weight of neurons in the first spatial pool and the second spatial pool, the connection between neurons is stable for normal continuous input, is not sensitive to abnormal input, has stronger robustness, and the reliability of the method is improved, and the spatial pool can also amplify differences, facilitate comparison between different contours, and improve the correct rate of identification;

[0123] 4、The embodiment of the application does not need to model the human body, simplifies the identification process, and improves the identification efficiency.

[0124] Reference Figure 2 The embodiment of the application also provides a gait recognition device based on a cerebral cortex algorithm, comprising:

[0125] A first module 601 is used for acquiring a gait image;

[0126] A second module 602 is used for performing gait segmentation on the gait image to obtain a binary contour image;

[0127] A third module 603 is used for encoding the binary contour image through a gait image to obtain an encoding result of gait segmentation binary image;

[0128] A fourth module 604 is used for performing sparse discrete representation on the encoding result of gait segmentation binary image through a gait image to obtain sparse representation of gait segmentation images;

[0129] A fifth module 605 is used for comparing the sparse representation of gait segmentation images with pre-stored sparse discrete representation of gait to obtain a gait recognition result.

[0130] The embodiment of the application also provides an electronic device, comprising a processor and a memory; the memory is used for storing a program; the processor executes the program to realize the gait recognition method based on the cerebral cortex algorithm.

[0131] The embodiment of the application also provides a computer readable storage medium, the storage medium stores a program, and the program is executed by a processor to realize the gait recognition method based on the cerebral cortex algorithm.

[0132] The embodiment of the present application further discloses a computer program product or computer program, which comprises computer instructions stored in a computer readable storage medium. A processor of a computer device can read the computer instructions from the computer readable storage medium, and the processor executes the computer instructions to enable the computer device to perform the method shown in the embodiment of the present application. Figure 1 The method shown in the embodiment of the present application.

[0133] In some alternative embodiments, the functions / operations mentioned in the block diagram can not occur in the order mentioned in the operation diagram. For example, depending on the functions / operations involved, two blocks shown in succession can actually be executed substantially simultaneously with each other, or the blocks can sometimes be executed in reverse order. In addition, the embodiments presented and described in the flowcharts of the present application are provided by way of example, and the purpose is to provide a more comprehensive understanding of the technology. The disclosed method is not limited to the operations and logical flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed and in which sub-operations described as part of larger operations are independently executed.

[0134] In addition, although the present application is described in the context of functional modules, it should be understood that, unless otherwise stated to the contrary, one or more of the described functions and / or features can be integrated in a single physical device and / or software module, or one or more functions and / or features can be implemented in separate physical devices or software modules. It can also be understood that a detailed discussion of the actual implementation of each module is unnecessary for an understanding of the present application. Rather, given the properties, functions and internal relationships of the various functional modules in the devices disclosed herein, the actual implementation of the module will be within the routine skill of the engineer, given the benefit of this disclosure. Thus, a person of ordinary skill in the art, using the ordinary skill, can implement the present application as set forth in the claims without undue experimentation. It can also be understood that the disclosed specific concepts are merely illustrative and are not intended to limit the scope of the present application, which is determined by the full scope of the appended claims and their equivalents.

[0135] If the functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the parts of the technical solutions that essentially contribute to the prior art or the parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0136] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a list of executable instructions for implementing logic functions, which can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus or device, such as a computer-based system, a system including a processor or other system that can fetch the instructions from the instruction execution system, apparatus or device and execute the instructions, or in conjunction with these instructions execution systems, apparatus or devices. For the purpose of this specification, the "computer-readable medium" can be any device that can contain, store, communicate, propagate or transport programs for use by or in connection with an instruction execution system, apparatus or device, or in conjunction with these instruction execution systems, apparatus or devices.

[0137] More specific examples (a non-exhaustive list) of the computer-readable medium include the following: an electrical connection having one or more wires (electrical devices), a portable computer diskette (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). In addition, the computer readable medium can even be paper or other suitable medium on which the program can be printed, because the program can be electronically obtained, for example, by optical scanning of the paper or other medium, followed by editing, interpreting or otherwise processing, if necessary, in other suitable ways, to be electronically obtained and then stored in the computer memory.

[0138] It should be understood that aspects of the application can be implemented in hardware, software, firmware or a combination thereof. In the above embodiments, various steps or methods can be implemented in software or firmware which is stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, and in another embodiment, any of the following technologies, known in the art, can be used: a combination of discrete logic circuits having logic gates for implementing logic functions upon an application of data signals, application specific integrated circuits having logic gates, field programmable gate arrays (FPGA), or other components, in combination or as the case can be.

[0139] In the description of the specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the application. In the specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in one or more embodiments or examples.

[0140] Although the embodiments of the present application have been shown and described, it would be appreciated by those skilled in the art that changes, modifications, alternatives and variations can be made to the embodiments without departing from the principles and spirit of the application, and the scope of the present application is defined by the claims and their equivalents.

[0141] The above is a specific description of the preferred embodiments of the present application, but the present application is not limited to the described embodiments, and those skilled in the art can make various equivalent modifications or replacements without departing from the spirit of the present application, and these equivalent modifications or replacements are all included in the scope defined by the claims of the present application.

Claims

1. A gait recognition method based on a brain cortex algorithm, characterized in that, The method comprises the following steps: obtaining a gait image; performing gait segmentation on the gait image to obtain a binary contour image; encoding the binary contour image to obtain an encoding result of the gait segmentation binary image; performing sparse discrete representation on the encoding result of the gait segmentation binary image to obtain a sparse representation of the gait segmentation image; comparing the sparse representation of the gait segmentation image with a pre-stored sparse discrete representation of the gait to obtain a gait recognition result; the step of encoding the binary contour image to obtain an encoding result of the gait segmentation binary image comprises: inputting the binary contour image into a first spatial pool to determine candidate neurons connected to the input data; increasing the weight of the candidate neurons to obtain activated neurons; updating the activated neurons to obtain the encoding result of the gait segmentation binary image.

2. The gait recognition method based on brain cortex algorithm according to claim 1, characterized in that, The step of performing gait segmentation on the gait image to obtain a binary contour image comprises: model solidification of an RVM algorithm model; setting input parameters of the RVM algorithm model after model solidification, the input parameters including a running device, a number of cycles, and a downsampling ratio; inputting the gait image into the RVM algorithm model to perform gait segmentation and obtain a binary contour image. 3.The gait recognition method based on brain cortex algorithm according to claim 1, characterized in that, The step of updating the activated neurons to obtain the encoding result of the gait segmentation binary image comprises: updating the weight of the activated neurons based on a first update formula; updating the reinforcement coefficient of the activated neurons based on a second update formula; updating the constraint radius of the activated neurons based on a third update formula.

4. The gait recognition method based on brain cortex algorithm according to claim 1, characterized in that, The step of performing sparse discrete representation on the encoding result of the gait segmentation binary image to obtain a sparse representation of the gait segmentation image comprises: adjusting the neuron activation rate of a second spatial pool; inputting the encoding result of the gait segmentation binary image into the adjusted second spatial pool to perform sparse discrete representation and obtain a sparse representation of the gait segmentation image.

5. The gait recognition method based on brain cortex algorithm according to claim 1, characterized in that, The step of comparing the sparse representation of the gait segmentation image with a pre-stored sparse discrete representation of the gait to obtain a gait recognition result comprises: storing the pre-stored sparse discrete representation of the gait in a dictionary form; performing an exhaustive comparison between the sparse representation of the gait segmentation image and the pre-stored sparse discrete representation of the gait to obtain a comparison similarity; comparing a pre-set comparison similarity threshold with the comparison similarity to obtain a gait recognition result; when the comparison similarity is greater than or equal to the pre-set comparison similarity threshold, it is determined that the target person is recognized; when the comparison similarity is less than the pre-set comparison similarity threshold, it is determined that the target person is not recognized.

6. The gait recognition method based on brain cortex algorithm according to claim 2, characterized in that, The step of inputting the gait image into the RVM algorithm model to perform gait segmentation and obtain a binary contour image comprises: performing feature extraction and semantic segmentation on the gait image through a feature encoder of the RVM algorithm model to obtain a first contour image; aggregating the temporal information of the first contour image through a recurrent decoder of the RVM algorithm model to obtain a second contour image; The second contour image is high-resolution predicted by a deep guided filter of the RVM algorithm model, to obtain a binary contour image.

7. A gait recognition device based on a brain cortex algorithm, characterized by, Comprise: A first module for acquiring a gait image; A second module for performing gait segmentation on the gait image to obtain a binary contour image; A third module for encoding the binary contour image to obtain an encoding result of the gait segmentation binary image; A fourth module for performing sparse discrete representation on the encoding result of the gait segmentation binary image to obtain a gait segmentation image sparse representation; A fifth module for comparing the gait segmentation image sparse representation with a pre-stored gait sparse discrete representation to obtain a gait recognition result; The third module is specifically used for: Inputting the binary contour image into a first spatial pool to determine candidate neurons connected to the input data; Adding weights to the candidate neurons to obtain activated neurons; Updating the activated neurons to obtain an encoding result of the gait segmentation binary image.

8. An electronic device, comprising: Comprise a processor and a memory; The memory is used to store a program; The processor executes the program to realize the method of any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The storage medium stores a program, and the program is executed by the processor to realize the method of any one of claims 1 to 6. The storage medium stores a program, and the program is executed by the processor to realize the method of any one of claims 1 to 6.

Images