Target detection tracking system and method based on artificial intelligence and machine vision

Abstract

The invention provides a target detection tracking system and method based on artificial intelligence and machine vision, and belongs to the technical field of image recognition. The target module is used for determining a monitored and tracked target object and determining a target application scene of the system; the data acquisition module is used for collecting original video data by using video acquisition equipment in a target application scene, and preprocessing the original video data to obtain a standard frame sequence; the labeling module is used for analyzing boundary features of the targets of all the standard frame sequences and identifying all the targets of the standard frame sequences to obtain labeling data; the model training module is used for training the initial model to obtain a final model; and the detection tracking module is used for inputting the real-time video stream into the final model, identifying the target of the video frame and tracking the position of the target in real time. And the accuracy and real-time performance of target identification and tracking are improved.

Description

Technical Field

[0001] The present invention relates to the field of image recognition technology, and in particular to a target detection and tracking system and method based on artificial intelligence and machine vision. Background Art

[0002] In recent years, the rapid development of deep learning technology has greatly improved the accuracy and efficiency of image processing and target recognition. The increasing demand for urban security and public place monitoring has promoted the application of target detection and tracking technology. However, due to different environmental conditions (such as lighting, weather, and occlusion), recognition technology may have errors and cause the loss of the target.

[0003] Therefore, the present invention proposes a target detection and tracking system and method based on artificial intelligence and machine vision. Summary of the invention

[0004] The present invention provides a target detection and tracking system and method based on artificial intelligence and machine vision, which are used to identify and mark targets in images after acquiring images using an image acquisition device, train a model, and detect and track targets in an actual environment based on the model.

[0005] In one aspect, the present invention provides a target detection and tracking system based on artificial intelligence and machine vision, comprising: Target module: defines the target objects to be monitored and tracked, and determines the target application scenarios of the system; Data acquisition module: Use video acquisition equipment to collect raw video data in the target application scenario, and pre-process the raw video data to obtain a standard frame sequence; Annotation module: analyzes the boundary features of all targets in the standard frame sequence, identifies all targets in the standard frame sequence, and obtains annotation data; Model training module: trains the initial model based on the labeled data and historical data in the history library to obtain the final model; Detection and tracking module: Input the real-time video stream into the final model, identify the target in the video frame, and track the target position in real time.

[0006] On the other hand, the target module comprises: Feature unit: defines the characteristics of the target object according to the target object to be monitored and tracked; Scenario unit: defines the monitoring and tracking environment of the target object and determines the target application scenario of the system.

[0007] On the other hand, the data acquisition module comprises: Corresponding unit: selecting a video acquisition device corresponding to the identification feature according to the feature of the target object, configuring a unique first number for the video acquisition device, and configuring a unique second number at the installation position of the video acquisition device in the target application scenario; An installation unit: installing the video acquisition device to the corresponding video acquisition device installation position according to a matching relationship between a first number of any video acquisition device and a second number of a corresponding video acquisition device installation position; Configuration unit: Determine the configuration parameters of the video acquisition device according to the target application scenario, configure and start all video acquisition devices, and obtain the original video data in real time.

[0008] On the other hand, the data acquisition module further includes: Video processing unit: unifies the data format of any original video data according to a preset standard, determines the predetermined number of frames according to a frequency-video duration-video rate mapping table, processes the original video frame by frame according to the predetermined number of frames to obtain multiple images, and obtains an original frame sequence; Frame processing unit: scaling any frame of the original frame sequence according to a bilinear method to make all frames of the original frame sequence have a unified standard size; standardizing the pixel value of any frame to obtain a standard frame; All standard frames in the original frame sequence constitute a standard frame sequence.

[0009] On the other hand, the marking module further includes: Edge acquisition unit: acquires any frame image of any standard frame sequence, performs Gaussian filtering to acquire an edge-clear image, and performs grayscale processing on each pixel point of the edge-clear image to form a grayscale image; Suppose the gray value of any pixel on the gray image is , get the absolute value of the L operator of any pixel point and perform threshold processing: ;in, Represents the current grayscale image edge L operator threshold, Represents the mean of the absolute values ​​of all L operators of the current grayscale image, Indicates that the current grayscale image has pixels, Represents the maximum absolute value of all L operators in the current grayscale image; Represents the minimum value of all L operators corresponding to the current grayscale image. Represents the absolute value of the L operator corresponding to the i-th pixel in the current grayscale image, Represents the horizontal grayscale value of the i-th pixel. Represents the vertical grayscale value of the i-th pixel; If the edge L operator of any pixel point is greater than the edge L operator threshold, the corresponding pixel point is an edge point; if the edge L operator of any pixel point is less than or equal to the edge L operator threshold, the corresponding pixel point is an internal point; Annotation unit: based on the closed image target formed by the edge points of any frame image, the feature name of the closed image target is annotated by using an annotation tool, all the closed image targets of the standard frame sequence are identified and constitute the image annotation part, and the annotation data is obtained, wherein the annotation data includes the image annotation part and the target motion change part.

[0010] On the other hand, the model training module includes: Training unit: divide the labeled data with the same feature name into a test set, and divide the historical data with the same feature name in the history library into a training set; Select the detection and tracking algorithms as the model architecture, where all image annotation parts are used as detection training, and all image target motion change parts are used as tracking training. Set the initial model parameters to build the initial model. The training set is used to train the initial model to obtain the first model, the test set is used to verify the first model to obtain the model parameter deviation, and the first model is adjusted based on the model parameter deviation to obtain the final model.

[0011] On the other hand, the detection and tracking module includes: Docking unit: deploys the final model into the actual environment and connects it to the real-time video stream interface; Monitoring unit: The final model analyzes and identifies the target of the real-time video stream in real time, tracks the analysis and tracking position of the real-time video stream in real time, compares it with the actual standard position, and updates and retrains the final model.

[0012] In another aspect, the present invention provides a target detection and tracking method based on artificial intelligence and machine vision, comprising: Step 1: Identify the target objects to be monitored and tracked, and determine the application scenarios of the system; Step 2: Use a video acquisition device to collect raw video data in the target application scenario, and preprocess the raw video data to obtain a standard frame sequence; Step 3: Analyze the boundary features of all targets in the standard frame sequence, identify all targets in the standard frame sequence, and obtain labeled data; Step 4: Train the initial model based on the labeled data and historical data from the history library to obtain the final model; Step 5: Input the real-time video stream into the first model, identify the target in the video frame, and track the target position in real time.

[0013] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a target detection and tracking system and method based on artificial intelligence and machine vision, which are used to identify and mark targets in images after acquiring images using an image acquisition device, train a model, and detect and track targets in an actual environment based on the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0015] Figure 1 It is a structural schematic diagram of a target detection and tracking system based on artificial intelligence and machine vision provided by an embodiment of the present invention; Figure 2 It is a flowchart of a target detection and tracking method based on artificial intelligence and machine vision provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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.

[0017] Embodiment 1: like Figure 1 As shown, an embodiment of the present invention provides a target detection and tracking system based on artificial intelligence and machine vision, including: Target module: defines the target objects to be monitored and tracked, and determines the target application scenarios of the system; Data acquisition module: Use video acquisition equipment to collect raw video data in the target application scenario, and pre-process the raw video data to obtain a standard frame sequence; Annotation module: analyzes the boundary features of all targets in the standard frame sequence, identifies all targets in the standard frame sequence, and obtains annotation data; Model training module: trains the initial model based on the labeled data and historical data in the history library to obtain the final model; Detection and tracking module: Input the real-time video stream into the final model, identify the target in the video frame, and track the target position in real time.

[0018] In this embodiment, the target object refers to a specific object that needs to be identified and tracked in the video monitoring and tracking system, such as a person, a vehicle, an animal, etc.

[0019] In this embodiment, target application scenarios include: public places, home environments, industrial environments, etc.

[0020] In this embodiment, the video acquisition device is a hardware tool for collecting raw video data, such as a camera, a drone, a mobile phone, etc.

[0021] In this embodiment, the original video data refers to the unprocessed original video content recorded by a video acquisition device (such as a camera) in the target application scenario.

[0022] In this embodiment, preprocessing is an important step in the data processing process, which aims to clean and convert the original video data.

[0023] In this embodiment, the standard frame sequence refers to frames with uniform format and features extracted from the original video data after preprocessing.

[0024] In this embodiment, the target analysis boundary features refer to features used to describe and identify the boundaries and shapes of target objects in an image or video frame, such as boundaries, contours, key points, etc.

[0025] In this embodiment, the annotation data refers to the information about the target in the image or video frame, including: target position, target type, time, etc.

[0026] In this embodiment, the history library refers to a database that stores past data and model training information.

[0027] In this embodiment, historical data refers to relevant data collected during past monitoring and tracking processes.

[0028] In this embodiment, the initial model refers to the first version of the model created based on the labeled data and historical data during the model training process.

[0029] In this embodiment, the final model is a machine learning or deep learning model that has been trained and optimized.

[0030] In this embodiment, the real-time video stream refers to a continuous image data stream generated by a video acquisition device (such as a camera).

[0031] In this embodiment, a video frame refers to a single still image in a video, and is a basic component unit of a video stream.

[0032] The working principle and beneficial effects of the above technical solution are: through clear targets, data collection, target labeling, model training and real-time detection and tracking, efficient identification and tracking of monitored objects can be achieved, thereby improving the system's monitoring capabilities and real-time response performance.

[0033] Embodiment 2: Based on the above embodiment 1, the target module includes: Feature unit: defines the characteristics of the target object according to the target object to be monitored and tracked; Scenario unit: defines the monitoring and tracking environment of the target object and determines the target application scenario of the system.

[0034] In this embodiment, features are key attributes used to describe the target object, including color, shape, texture, etc.

[0035] In this embodiment, the environment refers to the physical space where the target object exists and moves.

[0036] The working principle and beneficial effects of the above technical solution are: by defining target characteristics and monitoring environment, it ensures that the system can accurately identify and track target objects, improve monitoring efficiency, adapt to different application scenarios, and enhance the flexibility and accuracy of the system.

[0037] Embodiment 3: Based on the above embodiment 2, the data acquisition module includes: Corresponding unit: selecting a video acquisition device corresponding to the identification feature according to the feature of the target object, configuring a unique first number for the video acquisition device, and configuring a unique second number at the installation position of the video acquisition device in the target application scenario; An installation unit: installing the video acquisition device to the corresponding video acquisition device installation position according to a matching relationship between a first number of any video acquisition device and a second number of a corresponding video acquisition device installation position; Configuration unit: Determine the configuration parameters of the video acquisition device according to the target application scenario, configure and start all video acquisition devices, and obtain the original video data in real time.

[0038] In this embodiment, the first number refers to a unique identifier assigned to each video acquisition device.

[0039] In this embodiment, the second number refers to a unique identifier assigned to each installation location of the video acquisition device in the target application scenario.

[0040] In this embodiment, the matching relationship refers to the association between the unique identifier (first number) of the video acquisition device and the unique identifier (second number) of its installation location.

[0041] In this embodiment, the installation location refers to a specific place or area where a specific video acquisition device is installed in a target application scenario.

[0042] In this embodiment, the configuration parameters refer to various parameters and options required when determining the settings of the video acquisition device.

[0043] The working principle and beneficial effects of the above technical solution are: by matching target features with video equipment, ensuring the correct installation and configuration of the equipment, improving the accuracy of data collection, and obtaining high-quality original video in real time, laying the foundation for subsequent analysis.

[0044] Embodiment 4: Based on the above embodiment 3, the data acquisition module further includes: Video processing unit: unifies the data format of any original video data according to a preset standard, determines the predetermined number of frames according to a frequency-video duration-video rate mapping table, processes the original video frame by frame according to the predetermined number of frames to obtain multiple images, and obtains an original frame sequence; Frame processing unit: scaling any frame of the original frame sequence according to a bilinear method to make all frames of the original frame sequence have a unified standard size; standardizing the pixel value of any frame to obtain a standard frame; All standard frames in the original frame sequence constitute a standard frame sequence.

[0045] In this embodiment, the preset standard refers to a series of regulations and requirements for unifying the data format and processing the original video data during the video processing process, including: data format, resolution, frame rate, etc.

[0046] In this embodiment, the data format refers to the specific structure and encoding method used by video and image data during storage and transmission.

[0047] In this embodiment, the frequency-video duration-video rate mapping table is a table used to determine the relationship between the number of video frames, the video duration and the playback rate.

[0048] In this embodiment, the predetermined number of frames refers to the expected number of frames calculated according to parameters such as the total duration of the video, the playback rate, and the set frame rate during the video processing.

[0049] In this embodiment, frame-by-frame processing refers to a process of analyzing and processing each frame of the video separately.

[0050] In this embodiment, the bilinear method is an image scaling and interpolation technology, which obtains four neighboring pixels around the target pixel and generates a new pixel value by weighted calculation of the values ​​of the four pixels.

[0051] In this embodiment, the pixel value refers to the color and brightness information represented by each pixel in the image.

[0052] In this embodiment, the standardization process refers to a process of uniformly adjusting the pixel values ​​in an image so that different images have consistent brightness, contrast and color distribution.

[0053] In this embodiment, the standard frame refers to an image frame that meets a preset standard size and a standardized pixel value after being processed.

[0054] The working principle and beneficial effects of the above technical solution are: by unifying the video format, determining the predetermined number of frames and standardizing the processing frames, a consistent standard frame sequence is generated, the comparability and processing efficiency of the video data are improved, and a reliable basis is provided for subsequent analysis.

[0055] Embodiment 5: Based on the above embodiment 1, the marking module further includes: Edge acquisition unit: acquires any frame image of any standard frame sequence, performs Gaussian filtering to acquire an edge-clear image, and performs grayscale processing on each pixel point of the edge-clear image to form a grayscale image; Suppose the gray value of any pixel on the gray image is , get the absolute value of the L operator of any pixel point and perform threshold processing: ;in, Represents the current grayscale image edge L operator threshold, Represents the mean of the absolute values ​​of all L operators of the current grayscale image, Indicates that the current grayscale image has pixels, Represents the maximum absolute value of all L operators in the current grayscale image; Represents the minimum value of all L operators corresponding to the current grayscale image. Represents the absolute value of the L operator corresponding to the i-th pixel in the current grayscale image, Represents the horizontal grayscale value of the i-th pixel. Represents the vertical grayscale value of the i-th pixel; If the edge L operator of any pixel point is greater than the edge L operator threshold, the corresponding pixel point is an edge point; if the edge L operator of any pixel point is less than or equal to the edge L operator threshold, the corresponding pixel point is an internal point; Annotation unit: based on the closed image target formed by the edge points of any frame image, the feature name of the closed image target is annotated by using an annotation tool, all the closed image targets of the standard frame sequence are identified and constitute the image annotation part, and the annotation data is obtained, wherein the annotation data includes the image annotation part and the target motion change part.

[0056] In this embodiment, the Gaussian filtering process achieves a smoothing effect by performing a convolution operation on the image using a Gaussian function.

[0057] In this embodiment, the image with clear edge refers to an image that can obviously highlight the edge features of objects in the image after being processed.

[0058] In this embodiment, grayscale processing is a process of converting a color image into a grayscale image in image processing, and converting each pixel from color to a single color.

[0059] In this embodiment, the grayscale image refers to the result of converting a color image into a single-channel image, in which the value of each pixel represents its brightness or grayscale level.

[0060] In this embodiment, the grayscale value is a numerical value used to represent the brightness of each pixel in image processing.

[0061] In this embodiment, the L operator absolute value is an operator used for edge detection, and the edge is mainly detected by calculating the gradient of pixels in the image.

[0062] In this embodiment, the edge L operator threshold is a key parameter for determining edge points in an image, and the standard value of the absolute values ​​of all L operators in the image is obtained to represent the edge strength.

[0063] In this embodiment, the edge point refers to a position in the image where a pixel value changes significantly, and represents a boundary or shape feature of an object.

[0064] In this embodiment, the internal points refer to those pixel points that are not edges in the image.

[0065] In this embodiment, the closed image target refers to an object in the image that is identified by edge detection and annotation tools and forms a complete and closed outline or area.

[0066] In this embodiment, the target motion change portion refers to changes in displacement, rotation, scaling, etc. of the target in the image sequence over time.

[0067] The working principle and beneficial effects of the above technical solution are: extracting image edges through Gaussian filtering and grayscale processing, using L operator threshold to determine edge points, and then performing target labeling to generate labeling data containing feature names and motion changes, thereby improving the accuracy and practicality of image recognition.

[0068] Embodiment 6: Based on the above embodiment 5, the model training module includes: Training unit: divide the labeled data with the same feature name into a test set, and divide the historical data with the same feature name in the history library into a training set; Select the detection and tracking algorithms as the model architecture, where all image annotation parts are used as detection training, and all image target motion change parts are used as tracking training. Set the initial model parameters to build the initial model. The training set is used to train the initial model to obtain the first model, the test set is used to verify the first model to obtain the model parameter deviation, and the first model is adjusted based on the model parameter deviation to obtain the final model.

[0069] In this embodiment, the test set is a data set used to evaluate and verify the performance of the model.

[0070] In this embodiment, the training set is a data set used to train a machine learning model.

[0071] In this embodiment, model architecture construction refers to the configuration of the framework and components used to define the model structure in machine learning, including: algorithm selection, hierarchy, parameters, etc.

[0072] In this embodiment, the detection and tracking algorithm refers to computer vision technology used to process image and video data, such as SSD technology, feature tracking algorithm, etc.

[0073] In this embodiment, the initial model parameters refer to the initial values ​​set when the model is constructed, including: weights, learning rates, regularization parameters, etc.

[0074] In this embodiment, the first model refers to a model obtained by preliminary training based on training set data during the training process.

[0075] In this example, the deviation refers to the difference between the model's performance on the test set and the actual expectation.

[0076] The working principle and beneficial effects of the above technical solution are: by dividing the training set and the test set, using labeled data to train the detection and tracking model, adjusting the model parameters to optimize the performance, and ultimately improving the accuracy and robustness of target detection and tracking.

[0077] Embodiment 7: Based on the above embodiment 1, the detection and tracking module includes: Docking unit: deploys the final model into the actual environment and connects it to the real-time video stream interface; Monitoring unit: The final model analyzes and identifies the target of the real-time video stream in real time, tracks the analysis and tracking position of the real-time video stream in real time, compares it with the actual standard position, and updates and retrains the final model.

[0078] In this embodiment, the actual environment refers to the real-world scenario where the final model is deployed and run.

[0079] In this embodiment, the interface refers to a protocol or connection method for communication and data exchange between different systems, devices or components.

[0080] In this embodiment, the analyzed tracking position refers to the target position recognized by the model in the real-time video stream.

[0081] In this embodiment, the actual standard position refers to a predefined target position in a specific application scenario.

[0082] The working principle and beneficial effects of the above technical solution are: by connecting the final model with the real-time video stream, real-time recognition and tracking of the target can be achieved, and compared with the standard position, the model can be updated and retrained in time to improve the real-time and accuracy of the system.

[0083] Embodiment 8: like Figure 2 As shown, an embodiment of the present invention provides a target detection and tracking method based on artificial intelligence and machine vision, comprising: Step 1: Identify the target objects to be monitored and tracked, and determine the application scenarios of the system; Step 2: Use a video acquisition device to collect raw video data in the target application scenario, and preprocess the raw video data to obtain a standard frame sequence; Step 3: Analyze the boundary features of all targets in the standard frame sequence, identify all targets in the standard frame sequence, and obtain labeled data; Step 4: Train the initial model based on the labeled data and historical data from the history library to obtain the final model; Step 5: Input the real-time video stream into the first model, identify the target in the video frame, and track the target position in real time.

[0084] The working principle and beneficial effects of the above technical solution are: through clear targets, data collection, target labeling, model training and real-time detection and tracking, efficient identification and tracking of monitored objects can be achieved, thereby improving the system's monitoring capabilities and real-time response performance.

[0085] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A target detection and tracking system based on artificial intelligence and machine vision, characterized in that: include: Target module: defines the target objects to be monitored and tracked, and determines the target application scenarios of the system; Data acquisition module: Use video acquisition equipment to collect raw video data in the target application scenario, and pre-process the raw video data to obtain a standard frame sequence; Annotation module: analyzes the boundary features of all targets in the standard frame sequence, identifies all targets in the standard frame sequence, and obtains annotation data; Model training module: trains the initial model based on the labeled data and historical data in the history library to obtain the final model; Detection and tracking module: Input the real-time video stream into the final model, identify the target in the video frame, and track the target position in real time.

2. The target detection and tracking system based on artificial intelligence and machine vision according to claim 1, characterized in that: The target module comprises: Feature unit: defines the characteristics of the target object according to the target object to be monitored and tracked; Scenario unit: defines the monitoring and tracking environment of the target object and determines the target application scenario of the system.

3. The target detection and tracking system based on artificial intelligence and machine vision according to claim 2, characterized in that: The data acquisition module comprises: Corresponding unit: according to the characteristics of the target object, selects a video acquisition device corresponding to the identification characteristics, configures a unique first number for the video acquisition device, and configures a unique second number at the installation position of the video acquisition device in the target application scenario; An installation unit: installing the video acquisition device to the corresponding video acquisition device installation position according to a matching relationship between a first number of any video acquisition device and a second number of the corresponding video acquisition device installation position; Configuration unit: Determine the configuration parameters of the video acquisition device according to the target application scenario, configure and start all video acquisition devices, and obtain the original video data in real time.

4. The target detection and tracking system based on artificial intelligence and machine vision according to claim 3, characterized in that: The data acquisition module further includes: Video processing unit: unifies the data format of any original video data according to a preset standard, determines the predetermined number of frames according to a frequency-video duration-video rate mapping table, processes the original video frame by frame according to the predetermined number of frames to obtain multiple images, and obtains an original frame sequence; Frame processing unit: scaling any frame of the original frame sequence according to a bilinear method to make all frames of the original frame sequence have a unified standard size; standardizing the pixel value of any frame to obtain a standard frame; All standard frames in the original frame sequence constitute a standard frame sequence.

5. The target detection and tracking system based on artificial intelligence and machine vision according to claim 1, characterized in that: The marking module further includes: Edge acquisition unit: acquires any frame image of any standard frame sequence, performs Gaussian filtering to acquire an edge-clear image, and performs grayscale processing on each pixel point of the edge-clear image to form a grayscale image; Suppose the gray value of any pixel on the gray image is , get the absolute value of the L operator of any pixel point and perform threshold processing: ;in, Represents the current grayscale image edge L operator threshold, Represents the mean of the absolute values ​​of all L operators of the current grayscale image, Indicates that the current grayscale image has pixels, Represents the maximum absolute value of all L operators in the current grayscale image; Represents the minimum value of all L operators corresponding to the current grayscale image. Represents the absolute value of the L operator corresponding to the i-th pixel in the current grayscale image, Represents the horizontal grayscale value of the i-th pixel. Represents the vertical grayscale value of the i-th pixel; If the edge L operator of any pixel is greater than the edge L operator threshold, the corresponding pixel is an edge point; if the edge L operator of any pixel is less than or equal to the edge L operator threshold, the corresponding pixel is an internal point; Annotation unit: based on the closed image target formed by the edge points of any frame image, the feature name of the closed image target is annotated by using an annotation tool, all the closed image targets of the standard frame sequence are identified and constitute the image annotation part, and the annotation data is obtained, wherein the annotation data includes the image annotation part and the target motion change part.

6. The target detection and tracking system based on artificial intelligence and machine vision according to claim 5, characterized in that: The model training module includes: Training unit: divide the labeled data with the same feature name into a test set, and divide the historical data with the same feature name in the history library into a training set; Select the detection and tracking algorithms as the model architecture, where all image annotation parts are used as detection training, and all image target motion change parts are used as tracking training. Set the initial model parameters to build the initial model. The training set is used to train the initial model to obtain the first model, the test set is used to verify the first model to obtain the model parameter deviation, and the first model is adjusted based on the model parameter deviation to obtain the final model.

7. The target detection and tracking system based on artificial intelligence and machine vision according to claim 1, characterized in that: The detection and tracking module includes: Docking unit: deploys the final model into the actual environment and connects it to the real-time video stream interface; Monitoring unit: The final model analyzes and identifies the target of the real-time video stream in real time, tracks the analysis and tracking position of the real-time video stream in real time, compares it with the actual standard position, and updates and retrains the final model.

8. A target detection and tracking method based on artificial intelligence and machine vision, characterized in that: include: Step 1: Identify the target objects to be monitored and tracked, and determine the target application scenarios of the system; Step 2: Use a video acquisition device to collect raw video data in the target application scenario, and preprocess the raw video data to obtain a standard frame sequence; Step 3: Analyze the boundary features of all targets in the standard frame sequence, identify all targets in the standard frame sequence, and obtain labeled data; Step 4: Train the initial model based on the labeled data and historical data from the history library to obtain the final model; Step 5: Input the real-time video stream into the first model, identify the target in the video frame, and track the target position in real time.

Images