Low-cost gait recognition and suspicious behavior detection method

Abstract

The invention relates to a low-cost gait recognition and suspicious behavior detection method, and belongs to the field of image processing. According to the method, automatic identification and alarm of pedestrian behaviors are realized by using the existing monitoring camera and a deep learning algorithm. The method comprises the following specific steps: capturing a pedestrian target in a monitoring area in real time by using a camera, dynamically tracking the target, and continuously intercepting and storing pedestrian images in an anchor frame; performing edge detection processing on the stored image, and extracting main contour information of the image; n adjacent continuous images are selected to be stacked to form a three-dimensional tensor; and inputting the stacked images into a pre-trained classification system based on a convolutional neural network, analyzing behavior characteristics in an image sequence, judging whether a target behavior meets a suspicious behavior standard or not, and further determining a specific behavior category. The method has the advantages of low cost, high precision, strong real-time performance and the like.

Description

Technical Field

[0001] The invention belongs to the field of image processing and relates to a low-cost gait recognition and suspicious behavior detection method. Background Art

[0002] Suspicious behavior detection technology based on deep gait recognition has made significant progress in recent years. Its development status is mainly reflected in the following aspects:

[0003] First, the introduction of deep learning technology has greatly improved the accuracy and robustness of gait recognition. Through deep learning models such as convolutional neural networks (CNN) and long short-term memory networks (LSTM), gait features can be better extracted and analyzed, especially in complex backgrounds, lighting changes, occlusion and other environments. It can also maintain a high recognition accuracy. This makes gait recognition no longer limited to traditional simple feature extraction, and greatly enhances the ability to detect suspicious behavior.

[0004] Secondly, the construction of large-scale datasets provides important support for the training and evaluation of gait recognition models. Currently, multiple public gait datasets (such as CASIA-B, OU-ISIR, etc.) cover gait data in different environments, angles, and occlusion conditions, providing a good foundation for algorithm optimization and practical application testing. The continuous expansion and enrichment of these datasets enables the model to maintain strong generalization capabilities in a variety of complex scenarios.

[0005] Thirdly, multimodal fusion has become a development trend. In order to improve the accuracy of recognition and cope with the challenges of various environments, deep gait recognition is gradually combined with other biometric features (such as face and posture), and even integrates behavioral patterns in videos to construct a more comprehensive feature expression. This fusion method shows great potential in the fields of public security and intelligent monitoring, and can more accurately lock and analyze suspicious behaviors.

[0006] However, current technologies still face challenges in data privacy and ethical issues. The application of gait recognition involves continuous monitoring of individual behavior, which may raise concerns about privacy protection. In response, researchers are gradually exploring privacy protection solutions to ensure compliance with data usage.

[0007] The LiDAR-based gait recognition system uses LiDAR sensors to generate high-precision three-dimensional spatial data to analyze the gait characteristics of the human body. Its main technical points are as follows:

[0008] 3D point cloud data collection: LiDAR generates high-resolution 3D point cloud data by emitting laser pulses and receiving their reflected signals. This data can accurately reflect the position, shape and dynamic changes of the human body in space, thus supporting the collection of gait characteristics.

[0009] Point cloud preprocessing: The raw point cloud data generated by LiDAR usually contains noise and redundant information, so denoising, filtering and downsampling are required to improve data quality and computational efficiency. At the same time, point cloud segmentation technology is also used to extract the point cloud information of the target human body from the background.

[0010] Human posture modeling and tracking: By extracting geometric features from point cloud data, human posture modeling and tracking can be performed. Usually, joint point detection and skeleton reconstruction are used to generate a 3D human posture representation. These posture data provide more accurate dynamic features in gait recognition.

[0011] Gait feature extraction: Based on human posture, gait features are extracted using deep learning or geometric feature analysis methods. Gait features include step length, stride, joint angle changes, etc., which can effectively describe the dynamic characteristics of an individual. This process often uses convolutional neural networks (CNN) or graph convolutional networks (GCN) to learn and model data.

[0012] Behavior analysis and recognition: Utilize the extracted gait features for classification and recognition, and identify specific gait patterns through training models. Usually, algorithms such as support vector machines (SVM) and recurrent neural networks (RNN) are used to map different gait features to specific identities or behavior patterns to achieve gait recognition.

[0013] Multi-sensor fusion: In order to improve the adaptability and recognition accuracy of the system, LiDAR is often combined with other sensors (such as cameras and inertial sensors) to achieve multi-modal fusion of data. Multi-modal data can make up for the shortcomings of a single sensor and enhance the stability and robustness of the system in complex scenarios.

[0014] Real-time performance and computational optimization: The amount of data generated by LiDAR is large and the computation is complex, so real-time performance is crucial to the gait recognition system. The use of efficient point cloud processing algorithms, parallel computing technology (such as GPU acceleration) and optimized network structure can significantly improve the system's real-time response capability.

[0015] Through the implementation of the above technical points, the lidar-based gait recognition system can accurately capture and analyze human gait characteristics in a variety of scenarios, providing strong technical support for applications such as identity recognition and abnormal behavior monitoring.

[0016] The gait recognition system based on LiDAR has certain advantages in detection accuracy and data richness, but it also has some disadvantages and limitations:

[0017] High cost: LiDAR equipment is expensive, and the cost of building and maintaining the system is relatively high, especially when deployed on a large scale. This high cost will become a major obstacle to practical application.

[0018] Equipment size and complexity: LiDAR devices are usually large and complex, suitable for fixed monitoring scenarios, but lack portability and flexibility, so they are not easy to use in lightweight and highly mobile scenarios.

[0019] Environmental dependence: LiDAR systems perform well under certain conditions, but are easily disturbed in rainy and snowy weather, strong light, or environments with a large number of suspended particles in the air, affecting the accuracy and stability of the data.

[0020] Large amount of data processing: The amount of data collected by LiDAR is huge, especially when capturing complex movements or high-frequency sampling, which requires a lot of computing resources for data processing. This places high demands on the system hardware and may cause delays in real-time application scenarios.

[0021] Privacy and security issues: LiDAR technology can capture detailed three-dimensional information about the human body, which may involve personal privacy data. Although LiDAR does not directly capture image information, the high-precision three-dimensional data it generates may still bring challenges to privacy protection, especially in its widespread application in public places.

[0022] Although lidar has unique advantages in accuracy and applicability in gait recognition, its shortcomings in cost, equipment complexity, environmental adaptability and privacy issues need to be further addressed before it can be more widely used in gait recognition and suspicious behavior detection.

[0023] The present invention mainly solves the problems of high equipment cost, large size and complexity, and privacy concerns in the existing laser radar-based gait recognition system. By making full use of the existing camera-based monitoring system, combined with deep learning and computer vision technology, similar functions and similar performance indicators can be achieved at a lower cost. Summary of the invention

[0024] In view of this, the purpose of the present invention is to provide a low-cost gait recognition and suspicious behavior detection method, aiming to solve the shortcomings of the current gait recognition system based on laser radar, such as high equipment cost, large volume, complex structure and privacy protection. By making full use of the existing camera monitoring system and combining deep learning and computer vision technology, the function of abnormal behavior monitoring can be realized at a lower cost and achieve a performance level similar to that of the existing technology.

[0025] In order to achieve the above object, the present invention provides the following technical solutions:

[0026] The low-cost gait recognition and suspicious behavior detection method includes the following steps:

[0027] Use cameras to capture pedestrian targets in the monitoring area in real time;

[0028] Dynamically track pedestrian targets, and continuously capture and save pedestrian images in the anchor frame;

[0029] Perform edge detection on the saved image to extract the image contour information;

[0030] Select N adjacent continuous images and stack them to form a three-dimensional tensor;

[0031] The stacked images are input into a pre-trained classification system based on a convolutional neural network (CNN) to analyze the behavioral features in the image sequence;

[0032] Determine whether the target behavior meets the suspicious behavior criteria and further determine the specific behavior category;

[0033] If the system identifies suspicious behavior, it will output the identification results in real time and send an alarm to the security or monitoring system.

[0034] Furthermore, when dynamically tracking the pedestrian target and continuously capturing and saving the pedestrian images in the anchor frame, the YOLO (You Only Look Once: Unified, Real-Time Object Detection) target detection algorithm is used.

[0035] Furthermore, when edge detection is performed on the stored image, a Sobel operator or a Canny edge detection algorithm is used.

[0036] Furthermore, the suspicious behavior includes climbing over a fence.

[0037] Low-cost gait recognition and suspicious behavior detection system, the system includes:

[0038] A camera module is used to capture pedestrian targets in the monitoring area in real time;

[0039] The target detection module is used to dynamically track the target and continuously capture and save the pedestrian images in the anchor frame;

[0040] An image processing module is used to perform edge detection on the stored images and select N consecutive adjacent images for stacking;

[0041] The behavior recognition module is used to input the stacked images into a pre-trained classification system based on a convolutional neural network (CNN), analyze the behavior features in the image sequence, determine whether the target behavior meets the suspicious behavior criteria, and further determine the specific behavior category;

[0042] The alarm module is used to output the identification results in real time and send an alarm to the security or monitoring system if the system identifies suspicious behavior;

[0043] The camera module captures the pedestrian target images in the monitoring area in real time;

[0044] The target detection module uses the YOLO algorithm to detect and track pedestrian targets and capture pedestrian images within the anchor frame;

[0045] The image processing module performs edge detection on the captured pedestrian image and extracts image contour information;

[0046] The image stacking module stacks N consecutive edge-detected images to form a three-dimensional tensor;

[0047] The behavior recognition module inputs the stacked images into the CNN model, analyzes the behavior characteristics and performs behavior recognition;

[0048] If suspicious behavior is identified, the alarm module will output the identification results in real time and send an alarm to the security or monitoring system.

[0049] Furthermore, the target detection module adopts the YOLO target detection algorithm.

[0050] Furthermore, the image processing module adopts a Sobel operator or a Canny edge detection algorithm.

[0051] Furthermore, the behavior recognition module adopts a pre-trained convolutional neural network model.

[0052] The beneficial effect of the present invention is that by combining the YOLO target detection algorithm and the deep learning image classification model, the recognition accuracy and real-time performance similar to the laser radar solution are achieved in an economical and efficient manner. The specific steps include using a camera to capture pedestrian targets in real time, dynamic tracking and image edge processing, and realizing behavior analysis and automatic alarm output through image stacking and CNN classification.

[0053] Beneficial effects of the present invention

[0054] (1) Compared with the LiDAR-based system, this solution completely relies on the existing camera monitoring network and does not require expensive LiDAR equipment, reducing hardware investment. In addition, the system does not require complex installation and maintenance, further reducing labor and maintenance costs. Compared with traditional LiDAR, this solution achieves similar performance at a lower cost, reducing the economic pressure of system deployment and operation, and is particularly suitable for large-scale monitoring applications.

[0055] (2) Through the combination of the YOLO algorithm and the CNN classification model, this solution has high accuracy and reliability in capturing pedestrian targets and identifying behaviors. The YOLO target detection algorithm can efficiently identify and track pedestrians, while the CNN model analyzes pedestrian behavior through stacked images to accurately judge suspicious behaviors. Compared with traditional gait recognition, this solution can maintain stable recognition performance in complex environments, effectively improving the recognition efficiency and response speed of the monitoring system.

[0056] (3) The solution uses efficient YOLO detection and CNN classification models to process pedestrian behavior, which can track and identify targets in real time in the monitoring screen, and realize real-time monitoring and alarm output of suspicious behavior. Compared with traditional methods, this solution significantly improves the real-time performance and response speed of the system, providing strong support for the immediate handling of security incidents.

[0057] (4) This solution extracts gait features through edge detection, image stacking and other technologies, and combines it with the CNN deep learning model to achieve good adaptability to different lighting, occlusion and other environments, ensuring the behavior recognition effect in complex scenes. Compared with the lidar system, this solution is more applicable in more scenarios, reduces the false detection rate under specific weather conditions, and thus improves the overall reliability of the system.

[0058] (5) Unlike the precise spatial information capture of laser radar, this solution does not directly collect the three-dimensional contours of the human body, but only uses image data to complete gait recognition and behavior analysis, reducing the intrusion on personal privacy. At the same time, the data access control and encryption mechanism added to the system ensures the security of the data and meets the privacy protection needs in the monitoring field.

[0059] (6) This solution is highly scalable and can realize the combination and optimization and upgrading of multiple deep learning algorithms according to the needs of different application scenarios. Through modular design, the system can be easily expanded to other fields such as identity recognition and security monitoring, providing flexible solutions for different needs.

[0060] Traditional gait recognition based on LiDAR has problems such as high equipment cost, poor environmental adaptability, and high privacy risk. The present invention significantly reduces costs and privacy risks through a camera-based solution, and maintains high recognition accuracy and real-time performance in complex environments, effectively solving the shortcomings of the background technology.

[0061] The present invention aims to provide a low-cost, high-precision suspicious behavior detection method based on gait recognition, which can replace the gait recognition function of laser radar through camera system and deep learning, and maintain the same or similar detection effect. Ultimately, the present invention not only achieves the goals of low cost, high efficiency and privacy protection, but also provides higher adaptability, reliability and real-time performance, which meets the requirements of the invention purpose.

[0062] Other advantages, objectives and features of the present invention will be described in the following description to some extent, and to some extent, will be obvious to those skilled in the art based on the following examination and study, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below in conjunction with the accompanying drawings, wherein:

[0064] Figure 1 This is an example of a stacked record of pedestrians climbing over fences;

[0065] Figure 2 An example of a stacked record of pedestrian phone calls;

[0066] Figure 3 This is an example of a stacked image of pedestrian sitting and resting behavior records;

[0067] Figure 4 Example image after stacking for pedestrian standing behavior records;

[0068] Figure 5 An example of a stacked image of pedestrian walking behavior records. DETAILED DESCRIPTION

[0069] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments only illustrate the basic concept of the present invention in a schematic manner, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0070] Among them, the drawings are only used for illustrative explanations, and they only represent schematic diagrams rather than actual pictures, and should not be understood as limitations on the present invention. In order to better illustrate the embodiments of the present invention, some parts of the drawings may be omitted, enlarged or reduced, and do not represent the size of actual products. For those skilled in the art, it is understandable that some well-known structures and their descriptions in the drawings may be omitted.

[0071] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if the terms "upper", "lower", "left", "right", "front", "rear", etc. indicate the orientation or position relationship, they are based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation. Therefore, the terms describing the position relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting the present invention. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0072] The present invention proposes a low-cost camera-based gait recognition and suspicious behavior detection system. The system uses the existing camera monitoring network and combines deep learning algorithms to realize automatic recognition and alarm of pedestrian behavior.

[0073] System process:

[0074] 1. Target detection and capture: Use existing surveillance cameras and deep learning-based target recognition algorithms (such as YOLO) to detect and capture pedestrians in the surveillance area. Through YOLO's efficient anchor frame mechanism, pedestrian targets can be identified and marked in real time. The specific principles are as follows:

[0075] 1.1 Image Division

[0076] The input image is divided into S×S grids. Each grid is responsible for detecting the object in the area it covers.

[0077] 1.2 Bounding Box (Anchor Box) Prediction

[0078] Each grid predicts B bounding boxes. Each box contains the following four coordinates and a confidence score:

[0079] Position coordinates: center coordinates of the box (x, y), width w, height h.

[0080] Confidence score C: Confidence indicates the probability that this box contains the target, and the calculation formula is:

[0081] C=P(object)×IOU pred,truth

[0082] Among them, P(object) is the probability that the bounding box contains the object, IOU pred,truth Intersection over Union (IoU) of the predicted box and the true box

[0083] 1.3 Class Probability

[0084] The conditional probability of the target category predicted by each grid is the probability of belonging to a certain category when there is an object in this grid. For each category m, the probability is:

[0085] P(class m |object)

[0086] 1.4 Output Tensor

[0087] The final output is a tensor of S×S×(B×5+C), where: S×S is the number of grids, B×5 is the parameters of each bounding box (including position, width, height and confidence); C is the probability of each target category.

[0088] 1.5 Confidence Filtering

[0089] Only boxes that satisfy C>threshold are retained. This step ensures that boxes with lower confidence are filtered out, thus reducing noise.

[0090] 1.6 Non-Maximum Suppression (NMS)

[0091] Use non-maximum suppression (NMS) to remove duplicate boxes. For all predicted boxes of each category, calculate the IOU. If the IOU exceeds the specified threshold, only the box with the highest confidence is retained:

[0092] If IOU pred1,truth2 >Threshold, only boxes with higher confidence are retained

[0093] Through the above steps, YOLO can quickly and accurately detect pedestrian targets in images in one forward propagation.

[0094] 2 Dynamic tracking and image capture: After detecting the pedestrian target, the target is dynamically tracked, and the pedestrian image in the anchor frame is continuously captured and saved to ensure that the complete action process of the target is captured.

[0095] 2.1 Calculate the coordinates of the upper left and lower right corners of the bounding box

[0096] According to the anchor box related information obtained in process 1 (center coordinates (x, y), width w, height h of the box), the upper left corner (x min ,y min ) and the lower right corner coordinate (x max ,y max ) can be expressed as:

[0097]

[0098] 2.2 Capture the image area:

[0099] In the original image, the pixel range [x min :x max ,ymin :y max ] to capture the image. Assuming that the original image matrix is ​​Image, the sub-image in the capture box can be expressed as:

[0100] Image sub =Image[x min :x max ,y min :y max ]

[0101] Through the above formula, we can extract the image area within the bounding box from the original image.

[0102] 3. Edge detection and image stacking: Perform edge detection on the saved images to extract the main contour information of the images. Select N adjacent continuous images for stacking to capture the dynamic features of the target pedestrian on the time axis.

[0103] 3.1 Edge Detection Process

[0104] Image gradient or edge detection operators such as the Sobel operator or the Canny edge detection algorithm are usually used for edge detection. The following describes the edge detection process of a captured image using the Sobel operator as an example.

[0105] Let the captured image area be Image sub , whose pixel position is (i, j). Calculate the horizontal direction G at each pixel x and vertical direction G y gradient.

[0106]

[0107] The * here indicates the convolution operation.

[0108] Then use G x and G y Calculate the gradient magnitude (edge ​​strength) G at each pixel:

[0109]

[0110] In order to refine the edge, non-maximum suppression can be performed on G, and only the gradient position of the local maximum is retained as the edge.

[0111] By setting a threshold T, pixels with amplitude G greater than T are retained as edge points, and other pixels are set to 0:

[0112]

[0113] The obtained Edge is a binary image after edge detection, showing the edge in the captured image.

[0114] 3.2 Image stacking

[0115] N consecutive edge-detected images are stacked and represented as a three-dimensional tensor superposition process. Assume that the edge detection result of each image is recorded as Edge k , where k = 1, 2, ..., represents the kth frame image. The following is a description of the stacking process:

[0116] The edge detection result of each captured image k is recorded as the matrix Edge k (i, j), represents the edge information of the image at pixel (i, j).

[0117] N consecutive edge detection images are stacked into a three-dimensional tensor Stack(i,j,k), where the third dimension k represents the order of the time series. Specifically expressed as:

[0118] Stack(i,j,k)=Edge k (i,j)

[0119] The final stacking result:

[0120] Stack = {Edge1, Edge2, ..., Edge N}

[0121] 4. Gait behavior classification: The stacked images are input into a pre-trained classification system based on a convolutional neural network (CNN) to analyze the behavior characteristics in the image sequence. The system uses the classification model to determine whether the target behavior meets the suspicious behavior criteria and further determines the specific behavior category. The specific steps are as follows:

[0122] Input the input tensor Stack into the pre-trained CNN model. Assume that the mapping function of CNN is f CNN The forward propagation process is:

[0123] Feature = f CNN (Stack)

[0124] Among them, Feature is the feature vector extracted by CNN.

[0125] Input the feature vector Feature into the fully connected layer for classification. Assume that the mapping function of the fully connected layer is f FC And use the softmax activation function to output the category probability:

[0126] Class Probabilities=softmax(f FC(Feature))

[0127] Final predicted image category is the category corresponding to the maximum value in the category probability:

[0128] Output and alarm: If the system identifies suspicious behavior, it will output the identification results in real time and send an alarm to the security or monitoring system to facilitate timely response by relevant personnel.

[0129] The present invention integrates the existing camera monitoring system and the YOLO algorithm, and uses the CNN-based image stacking classification method to achieve low-cost and efficient gait recognition and suspicious behavior detection. The entire system has efficient data processing and accurate real-time performance, ensuring stable recognition accuracy in different environments.

[0130] YOLO-based target detection: Use the YOLO algorithm to detect pedestrian targets in real time in the monitoring screen, and combine it with the anchor frame mechanism to accurately capture and track pedestrians.

[0131] Image edge processing and stacking: Perform edge detection and stacking on continuous images to display pedestrian dynamic features on the time axis, providing key features for subsequent gait recognition and behavior judgment.

[0132] CNN classification model identifies behaviors: The stacked images are input into the pre-trained CNN model for classification, accurately determining whether the target behavior is in the "suspicious" category and identifying specific behavior patterns.

[0133] Real-time alarm mechanism: Once suspicious behavior is detected, the system automatically issues an alarm and outputs the results to help monitoring personnel respond quickly.

[0134] Efficient use of existing equipment: The present invention does not require expensive equipment such as lidar, and only implements gait recognition and anomaly detection based on existing cameras and deep learning algorithms, effectively reducing costs and simplifying deployment.

[0135] Highly adaptable multi-scenario applications: The system has high applicability and can operate stably in a variety of complex environments. It is suitable for security, public monitoring and other fields.

[0136] Through the implementation of the above technical points, the present invention can efficiently and economically complete gait recognition and suspicious behavior detection, and has broad application prospects in public safety and intelligent security.

[0137] In order to evaluate the behavior recognition method proposed in the present invention, we used 5 types of pedestrian behavior records extracted from public videos on the Internet through the method described in the technical solution and the main points of the invention, including 4 normal types of behaviors (standing, walking, making a phone call, sitting and resting), and 1 abnormal behavior (climbing over a fence) and adopted a cross-validation method. Figure 1 to Figure 5 Here is an example showing the various behaviors (these examples are the result of stacking the edge detected images):

[0138] We selected three CNN models that were pre-trained using different strategies for behavior recognition testing. The specific differences in training strategies are shown in Table 1:

[0139] Table 1

[0140]

[0141] Finally, the aggregate confusion matrix of behavior recognition in the test scenario is shown in Tables 2 to 4. Different tables represent the recognition test results of different pre-trained CNN models. The value of the mth row and nth column in the matrix represents the number of samples in the mth activity that are identified as the nth activity. This table shows the actual accuracy information of this method. Table 2 shows the overall average accuracy of CNN_1 recognition, 98.74%; Table 3 shows the overall average accuracy of CNN_2, 97.48%; Table 4 shows the overall average accuracy of CNN_3, 98.11%;

[0142] Table 2

[0143]

[0144] Table 3

[0145]

[0146] Table 4

[0147]

[0148]

[0149] Suspicious behavior categories (fence climbing) and most normal behaviors are almost always correctly identified, while walking behavior has a lot of false positives, presumably because small hand movements are sometimes difficult to distinguish due to occlusion, which makes walking and calling easy to confuse in some cases. It is also worth noting that the overall number of false positives is quite low (i.e., less than 5% of true behaviors), which indicates that the system rarely issues unexpected alarms when detecting pedestrian behavior types.

[0150] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution, which should be included in the scope of the claims of the present invention.

Claims

1. Low-cost gait recognition and suspicious behavior detection method, characterized by: The following steps are involved: Use cameras to capture pedestrian targets in the monitoring area in real time; Dynamically track pedestrian targets, and continuously capture and save pedestrian images in the anchor frame; Perform edge detection on the saved image to extract the image contour information; Select N adjacent continuous images and stack them to form a three-dimensional tensor; The stacked images are input into a pre-trained classification system based on a convolutional neural network (CNN) to analyze the behavioral features in the image sequence; Determine whether the target behavior meets the suspicious behavior criteria and further determine the specific behavior category; If the system identifies suspicious behavior, it will output the identification results in real time and send an alarm to the security or monitoring system.

2. The low-cost gait recognition and suspicious behavior detection method according to claim 1, characterized in that: The YOLO target detection algorithm is used when dynamically tracking pedestrian targets and continuously capturing and saving pedestrian images in anchor frames.

3. The low-cost gait recognition and suspicious behavior detection method according to claim 1, characterized in that: When performing edge detection on the stored image, a Sobel operator or a Canny edge detection algorithm is used.

4. The low-cost gait recognition and suspicious behavior detection method according to claim 1, characterized in that: The suspicious behavior included climbing over a fence.

5. Low-cost gait recognition and suspicious behavior detection system, characterized by: The system includes: A camera module is used to capture pedestrian targets in the monitoring area in real time; The target detection module is used to dynamically track the target and continuously capture and save the pedestrian images in the anchor frame; The image processing module is used to perform edge detection on the stored images and select N consecutive adjacent images for stacking; The behavior recognition module is used to input the stacked images into a pre-trained classification system based on a convolutional neural network (CNN), analyze the behavior features in the image sequence, determine whether the target behavior meets the suspicious behavior criteria, and further determine the specific behavior category; The alarm module is used to output the identification results in real time and send an alarm to the security or monitoring system if the system identifies suspicious behavior; The camera module captures the pedestrian target images in the monitoring area in real time; The target detection module uses the YOLO algorithm to detect and track pedestrian targets and capture pedestrian images within the anchor frame; The image processing module performs edge detection on the captured pedestrian image and extracts image contour information; The image stacking module stacks N consecutive edge-detected images to form a three-dimensional tensor; The behavior recognition module inputs the stacked images into the CNN model, analyzes the behavior characteristics and performs behavior recognition; If suspicious behavior is identified, the alarm module will output the identification results in real time and send an alarm to the security or monitoring system.

6. The low-cost gait recognition and suspicious behavior detection system according to claim 5, characterized in that: The target detection module adopts the YOLO target detection algorithm.

7. The low-cost gait recognition and suspicious behavior detection system according to claim 5, characterized in that: The image processing module adopts the Sobel operator or the Canny edge detection algorithm.

8. The low-cost gait recognition and suspicious behavior detection system according to claim 5, characterized in that: The behavior recognition module adopts a pre-trained convolutional neural network model.

Images