Video-based behavior track identification method

Abstract

The invention discloses a video-based behavior track identification method, and relates to the technical field of behavior track identification methods. A video-based behavior track identification method comprises a video behavior track identification method, a behavior identification module and a result display and application module. The data preprocessing module comprises a preprocessing module, a multi-scale analysis module, a distributed calculation module, an edge calculation module and a data storage module; by setting distributed calculation and edge calculation and distributing calculation tasks to a plurality of calculation nodes for parallel processing, the data processing speed and efficiency are remarkably improved, the single-point bottleneck is avoided, and the processing time of a large-scale video data set is greatly shortened; part of calculation tasks are processed on edge equipment for data acquisition, so that the bandwidth requirement and delay of data transmission are reduced, and the real-time performance and response speed of the system are improved; the behavior track of the target object in the video can be efficiently and accurately identified and tracked.

Description

Technical Field

[0001] The present invention relates to the technical field of behavior trajectory recognition methods, and in particular to a video-based behavior trajectory recognition method. Background Art

[0002] The video-based behavior trajectory recognition method obtains video data through video acquisition devices such as cameras, and then uses image processing, computer vision, artificial intelligence and other technologies to identify, track and analyze the target in the video to obtain its behavior trajectory. The system has wide application value in security monitoring, urban management, traffic management and other fields.

[0003] In the prior art, there is a method for identifying the trajectory of a drunk driving vehicle based on video surveillance, which includes a video acquisition module, a digital image preprocessing module, a vehicle driving feature detection module, a drunk driving vehicle identification module, a license plate recognition module, and a manual detection module for reporting to the traffic police, and each module is connected in sequence; wherein the digital image preprocessing module includes a lane line recognition module, a car identification module, and an image preprocessing module; the vehicle driving feature detection module includes a car tracking module, a turn signal recognition module, an angle detection module, and a line pressure detection module. The present invention expands the scope of drunk driving detection, improves the efficiency of the traffic police's drunk driving detection work, and makes drunk driving drivers no longer have a fluke mentality, thereby reducing the occurrence of traffic accidents caused by drunk driving.

[0004] However, the existing video behavior trajectory recognition method processes data at a single bottleneck, which greatly reduces the speed and efficiency of data processing. Therefore, a video-based behavior trajectory recognition method is needed. Summary of the invention

[0005] The present invention provides a behavior trajectory recognition method based on video to solve the problems in the background technology.

[0006] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: a video-based behavior trajectory recognition method, including a video behavior trajectory recognition method, the video behavior trajectory recognition method including a video acquisition module, a data preprocessing module, a target detection and tracking module, a feature extraction module, a trajectory generation module, a behavior recognition module and a result display and application module; the data preprocessing module includes a preprocessing module, multi-scale analysis, distributed computing, edge computing and data storage.

[0007] Furthermore, the preprocessing module is responsible for performing preliminary processing on the original video data, including denoising, normalization, and format conversion.

[0008] Furthermore, the multi-scale analysis module analyzes the video data at different scales.

[0009] Furthermore, the edge computing module can place part of the computing tasks on the edge device for data collection for processing, reducing the bandwidth requirements and delays of data transmission.

[0010] Furthermore, the distributed computing module can significantly improve the speed and efficiency of data processing by allocating computing tasks to multiple computing nodes for parallel processing. Distributed computing can also avoid single-point bottlenecks and greatly shorten the processing time for large-scale video data sets.

[0011] Furthermore, the data storage module is responsible for efficiently storing the preprocessed data so that subsequent modules can perform further analysis and processing.

[0012] Furthermore, the target detection and tracking module is divided into target detection and target tracking. Target detection: using a deep learning model or other algorithm to detect the target object in the video; target tracking: using a tracking algorithm to track the detected target and generate trajectory information, thereby locating and tracking the target object in the video.

[0013] Furthermore, the functions of the video acquisition module, data preprocessing module, feature extraction module, behavior recognition module and result display and application module are as follows: Video acquisition module: Use a camera or existing video file as the input source to ensure that the quality and resolution of the video meet the requirements of subsequent processing. If necessary, you can set the acquisition frequency and time period to obtain video data; Feature extraction module: Use a deep learning network to extract image features, use a time series analysis method to extract motion features, and combine feature extraction methods to obtain more comprehensive feature information, and extract features from the target object that are helpful for behavior recognition; Behavior recognition module: Use machine learning or deep learning models to analyze the extracted features, identify the target's behavior patterns, and identify specific behavior trajectories; Result display and application module: Display the identified behavior trajectory results to the user in a visual manner, or mark the target trajectory and behavior labels on the video, provide an application interface, and integrate the recognition results into other systems, such as security systems, intelligent monitoring systems, etc., to achieve automated response and decision support.

[0014] Compared with the prior art, the present invention provides a method for identifying behavior trajectories based on video, which has the following beneficial effects:

[0015] 1. This video-based behavior trajectory recognition method, by setting up distributed computing and edge computing, distributes computing tasks to multiple computing nodes for parallel processing, significantly improves the speed and efficiency of data processing, avoids single-point bottlenecks, and greatly shortens the processing time of large-scale video data sets; places some computing tasks on the edge devices for data acquisition, reduces the bandwidth requirements and delays for data transmission, and improves the real-time performance and response speed of the system; it can efficiently and accurately identify and track the behavior trajectory of target objects in the video.

[0016] 2. This video-based behavior trajectory recognition method can accurately calculate and generate motion trajectories by setting up a trajectory generation module, ensuring that the position, speed, and acceleration parameters during the execution process meet the predetermined requirements, thereby improving the accuracy of task execution; through algorithm optimization, the trajectory generation module can find the shortest or optimal path, reduce unnecessary movements, and improve task execution efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a structural diagram of the video behavior trajectory recognition method of the present invention;

[0018] Figure 2 This is a structural diagram of the data preprocessing module of the present invention.

[0019] In the figure: 1. Video behavior trajectory recognition method; 2. Video acquisition module; 3. Data preprocessing module; 4. Target detection and tracking module; 5. Feature extraction module; 6. Behavior recognition module; 7. Result display and application module; 301. Data storage; 302. Preprocessing module; 303. Multi-scale analysis; 304. Distributed computing; 305. Edge computing; 8. Trajectory generation module. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] See also Figure 1-2The present invention discloses a behavior trajectory recognition method based on video, including a video behavior trajectory recognition method 1, wherein the video behavior trajectory recognition method 1 includes a video acquisition module 2, a data preprocessing module 3, a target detection and tracking module 4, a feature extraction module 5, a trajectory generation module 8, a behavior recognition module 6 and a result display and application module 7; the data preprocessing module 3 includes a preprocessing module 302, a multi-scale analysis 303, a distributed computing 304, an edge computing 305 and a data storage 301.

[0022] Specifically, the pre-processing module 302 is responsible for performing preliminary processing on the original video data, including denoising, normalization, and format conversion.

[0023] In this embodiment, the pre-processing module 302 can improve data quality, reduce the impact of noise on subsequent analysis, and ensure that the data processed by subsequent modules are consistent and analyzable.

[0024] Specifically, the multi-scale analysis 303 module analyzes the video data at different scales.

[0025] In this implementation, the multi-scale analysis 303 module is capable of capturing behavioral features of different sizes and speeds, which helps to improve the accuracy and robustness of behavioral trajectory recognition, because behaviors may exhibit different features at different scales.

[0026] Specifically, the edge computing 305 module can place part of the computing tasks on the edge device for data collection for processing, reducing the bandwidth requirements and delays for data transmission.

[0027] Specifically, the distributed computing 304 module can significantly improve the speed and efficiency of data processing by allocating computing tasks to multiple computing nodes for parallel processing. Distributed computing 304 can also avoid single-point bottlenecks and greatly shorten the processing time for large-scale video data sets.

[0028] In this implementation, the edge computing 305 module is capable of responding quickly close to the data source.

[0029] Specifically, the data storage 301 module is responsible for efficiently storing the preprocessed data so that subsequent modules can perform further analysis and processing.

[0030] In this embodiment, an effective data storage 301 strategy can ensure the security and availability of data while supporting efficient query and access.

[0031] Specifically, the target detection and tracking module 4 is divided into target detection and target tracking. Target detection: using a deep learning model or other algorithms to detect target objects in the video; target tracking: using a tracking algorithm to track the detected target and generate trajectory information, thereby locating and tracking the target object in the video.

[0032] Specifically, the functions of the video acquisition module 2, the data preprocessing module 3, the feature extraction module 5, the behavior recognition module 6 and the result display and application module 7 are as follows: Video acquisition module 2: Use a camera or an existing video file as the input source to ensure that the quality and resolution of the video meet the requirements of subsequent processing. If necessary, you can set the acquisition frequency and time period to obtain video data; Feature extraction module 5: Use a deep learning network to extract image features, use a time series analysis method to extract motion features, and combine feature extraction methods to obtain more comprehensive feature information, and extract features from the target object that are helpful for behavior recognition; Behavior recognition module 6: Use machine learning or deep learning models to analyze the extracted features, identify the target's behavior patterns, and identify specific behavior trajectories; Result display and application module 7: Display the identified behavior trajectory results to the user in a visual way, or mark the target trajectory and behavior labels on the video, provide an application interface, and integrate the recognition results into other systems, such as security systems, intelligent monitoring systems, etc., to achieve automated response and decision support.

[0033] When in use, the video source can be collected from various devices such as surveillance cameras, mobile phones, and drones through the video acquisition module 2; then, the data preprocessing module 3 removes noise in the video, such as image noise and electronic noise; at the same time, adjusts the contrast of the video image to make the target object clearer; then, the multi-scale analysis 303 adjusts the brightness according to the video environment to ensure image quality; analyzes the video at different scales to capture target features at different distances and viewing angles, and then uses distributed computing 304 and edge computing 305 to accelerate data processing and reduce data transmission delays and bandwidth requirements; stores the preprocessed video data in a reliable data storage 301; models the video background through the target detection and tracking module 4 to distinguish between foreground target objects and backgrounds, analyze pixel changes between consecutive frames, and calculate motion vector fields to detect the motion trajectory of the target; extracts the color, texture, etc. of the target object Appearance features are used for subsequent tracking and identification; the above features are used to track the target object in continuous frames and record its position information. Feature extraction module 5: According to the motion trajectory, appearance features, etc. of the target object, a feature vector that can describe its behavior pattern is extracted; through trajectory generation module 8: the position information of the target object in each frame is connected to form a continuous motion trajectory; through behavior recognition module 6: if there is labeled behavior data in advance, a supervised learning method can be used to train the classifier to classify the behavior of new video data. If there is no labeled data, an unsupervised learning method can be used to mine the implicit rules in the behavior data to realize automatic recognition of behavior; using deep neural networks, feature representation and classification rules are automatically learned to improve the accuracy and robustness of behavior recognition. When abnormal behavior is detected, the relevant video screen is automatically displayed on the large screen through the result display and application module 7 to attract attention.

[0034] In summary, the video-based behavior trajectory recognition method, by setting up distributed computing 304 and edge computing 305, distributes computing tasks to multiple computing nodes for parallel processing, significantly improves the speed and efficiency of data processing, avoids single-point bottlenecks, and greatly shortens the processing time of large-scale video data sets; places part of the computing tasks on the edge device for data acquisition for processing, reduces the bandwidth requirements and delays for data transmission, and improves the real-time and response speed of the system; can efficiently and accurately identify and track the behavior trajectory of the target object in the video; by setting up the trajectory generation module 8, it can accurately calculate and generate the motion trajectory, ensuring that the position, speed, and acceleration parameters during the execution process meet the predetermined requirements, thereby improving the accuracy of task execution; through algorithm optimization, the trajectory generation module 8 can find the shortest or optimal path, reduce unnecessary movement, and improve task execution efficiency.

[0035] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A video-based behavior trajectory recognition method, comprising a video behavior trajectory recognition method (1), characterized in that: The video behavior trajectory recognition method (1) includes a video acquisition module (2), a data preprocessing module (3), a target detection and tracking module (4), a feature extraction module (5), a trajectory generation module (8), a behavior recognition module (6) and a result display and application module (7); the data preprocessing module (3) includes a preprocessing module (302), multi-scale analysis (303), distributed computing (304), edge computing (305) and data storage (301).

2. The method for video-based behavior trajectory recognition according to claim 1, characterized in that: The pre-processing module (302) is responsible for performing preliminary processing on the original video data, including denoising, normalization, and format conversion.

3. The method for video-based behavior trajectory recognition according to claim 1, characterized in that: The multi-scale analysis (303) module analyzes the video data at different scales.

4. The method for identifying behavior trajectories based on video according to claim 1, characterized in that: The edge computing (305) module can place part of the computing tasks on the edge device for data collection for processing, reducing the bandwidth requirements and delays of data transmission.

5. The method for video-based behavior trajectory recognition according to claim 1, characterized in that: The distributed computing (304) module can significantly improve the speed and efficiency of data processing by allocating computing tasks to multiple computing nodes for parallel processing. Distributed computing (304) can also avoid single-point bottlenecks and greatly shorten the processing time for large-scale video data sets.

6. The method for video-based behavior trajectory recognition according to claim 1, characterized in that: The data storage (301) module is responsible for efficiently storing the pre-processed data so that subsequent modules can perform further analysis and processing.

7. The method for video-based behavior trajectory recognition according to claim 1, characterized in that: The target detection and tracking module (4) is divided into target detection and target tracking. Target detection: using a deep learning model or other algorithms to detect target objects in the video; Target tracking: Use tracking algorithms to track detected targets and generate trajectory information, thereby locating and tracking target objects in the video.

8. The method for video-based behavior trajectory recognition according to claim 1, characterized in that: The functions of the video acquisition module (2), the data preprocessing module (3), the feature extraction module (5), the behavior recognition module (6) and the result display and application module (7) are as follows: the video acquisition module (2) uses a camera or an existing video file as an input source to ensure that the quality and resolution of the video meet the requirements of subsequent processing. If necessary, the acquisition frequency and time period can be set to obtain video data; the feature extraction module (5) uses a deep learning network to extract image features, uses a time series analysis method to extract motion features, and combines feature extraction methods to obtain more comprehensive feature information, and extracts features that are helpful for behavior recognition from the target object; Behavior recognition module (6): uses machine learning or deep learning models to analyze the extracted features and identify the target's behavior patterns, which can identify specific behavior trajectories; Result display and application module (7): Displays the identified behavior trajectory results to the user in a visual manner, or marks the target trajectory and behavior labels on the video, provides an application interface, and integrates the identification results into other systems, such as security systems, intelligent monitoring systems, etc., to achieve automated response and decision support.

Images