A tracking method and system based on intelligent big data

Abstract

The present invention provides a tracking method and system based on intelligent big data, which relates to the technical field of data processing. The method comprises: acquiring real-time image data, scene environment parameter data and auxiliary behavior data of a target object; processing the data to generate key frame data, dynamic environment characteristic data, static background characteristic data and behavior characteristic data; calculating the initial position and motion trend of the target object according to the key frame data and the behavior characteristic data in combination with the dynamic environment characteristics, and generating motion prediction data; dividing the activity range of the target object according to the motion prediction data and the static background characteristics, and generating target area planning data; formulating shooting path data according to the target area planning data, and generating a preliminary shooting plan; refining the preliminary shooting plan into a step-by-step shooting instruction set within a time period; storing the continuously shot pictures and related dynamic data as an expandable multidimensional data set; the present invention improves the intelligent effect of tracking.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a tracking method and system based on intelligent big data. Background Art

[0002] With the rapid development of big data technology, automated tracking systems based on big data analysis are gradually being applied to scenarios such as film and television production, live broadcasts of sports events, and public safety. Existing technologies usually achieve real-time tracking and shooting of target objects by deploying fixed cameras or using track cameras. Such systems use target detection and tracking algorithms to analyze pixel changes, target features, and trajectory information in video streams to generate corresponding control instructions to adjust the camera's shooting angle, focal length, and moving direction, thereby completing continuous shooting of the target object. However, this technology mainly relies on pre-set fixed algorithms and a single data source, making it difficult to dynamically adjust shooting strategies in complex environments.

[0003] There are technical defects in the application of existing technologies in real-time live broadcasts of sports events. For example, in football matches, existing automated tracking systems often have difficulty capturing fast-moving players and footballs at the same time. When multiple targets, such as players, footballs, and referees, move in a complex trajectory, the system is prone to misjudgment or loss of key images. This is because existing algorithms are usually rule-driven and lack comprehensive perception of dynamic changes in the environment and big data support, resulting in the loss of key details during the shooting process, and unable to meet the high requirements for picture quality and integrity of live broadcasts of events. Summary of the invention

[0004] The purpose of the present invention is to provide a tracking method and system based on intelligent big data, aiming to solve the problems mentioned in the background technology.

[0005] In order to solve the above technical problems, the technical solution of the present invention is as follows:

[0006] In a first aspect, a tracking method based on smart big data is provided, the method comprising:

[0007] Obtain real-time image data, scene environment parameter data, and auxiliary behavior data of the target object;

[0008] Perform frame processing on real-time image data, extract frames with significant changes in the picture, and generate key frame data;

[0009] Classify and analyze scene environment parameter data to extract dynamic environment characteristic data and static background characteristic data;

[0010] Analyze the auxiliary behavior data, extract the movement pattern of the target object, and generate behavior characteristic data;

[0011] According to the key frame data and behavior characteristic data, combined with the dynamic environment characteristics, the preliminary position and motion trend of the target object are calculated to generate motion prediction data;

[0012] According to the motion prediction data and static background characteristics, the activity range of the target object is divided and the target area planning data is generated;

[0013] According to the target area planning data, the shooting path data is formulated to generate a preliminary shooting plan, which includes the movement trajectory of the camera equipment, the shooting angle and the focal length change;

[0014] Refine the preliminary shooting plan into a step-by-step shooting instruction set within a time period, and dynamically adjust the shooting instructions of the step-by-step shooting instruction set according to the real-time changes in the behavior of the target object;

[0015] According to the step-by-step shooting instruction set, it is executed in real time, and the device is controlled to obtain continuous shooting images;

[0016] Real-time analysis of the captured images, evaluation of their quality, and generation of image quality feedback data;

[0017] Optimizing the shooting instructions of the step-by-step shooting instruction set according to the image quality feedback data to obtain an optimized shooting instruction set;

[0018] Store continuous shooting images and related dynamic data as expandable multi-dimensional datasets.

[0019] Preferably, the auxiliary behavior data is parsed to extract the movement pattern of the target object and generate behavior characteristic data, including:

[0020] Analyze auxiliary behavior data, identify the behavior frequency and change range of the target object, and extract behavior characteristics;

[0021] According to the time sequence information of the target object, the movement direction, speed and trajectory form in the behavior characteristics are analyzed to generate the movement pattern of the target object;

[0022] Combine the target object's movement pattern and behavioral characteristics to form behavioral characteristic data.

[0023] Preferably, the preliminary position and motion trend of the target object are calculated based on the key frame data and the behavior characteristic data in combination with the dynamic environment characteristics to generate motion prediction data, including:

[0024] Synchronously analyze the key frame data and the behavior characteristic data, and generate preliminary position distribution data by combining the position distribution and motion pattern of the target object;

[0025] According to the preliminary position distribution data and the dynamic environment characteristics in the environment characteristic data, the motion vector of the target object in different time periods is calculated to generate motion trend data; the motion vector is calculated by the following formula;

[0026] ,

[0027] in, For the target object at time The motion vector of For the target object at time point The motion vector of is the time step, For in time The total number of dynamic environments at the target object's location, For in time The number of target object behavior characteristics, For the The weight of the dynamic environment characteristics, For the The weight of the behavioral characteristics, For the The impact vector of dynamic environment characteristics on the target object, , For the target object at time The position vector of For the The position vector of the dynamic environment characteristic, For the The influence vector of the behavior characteristics on the target object, , For the The position vector of the behavioral characteristics, and is the magnitude of the corresponding vector;

[0028] Perform correlation analysis on the preliminary location distribution data and motion trend data, eliminate abnormal motion paths that are inconsistent with the dynamic environment, and generate location prediction data;

[0029] The position prediction data is divided into time periods to form a target position sequence of multiple time periods and generate motion prediction data.

[0030] Preferably, the activity range of the target object is divided according to the motion prediction data and the static background characteristics, and the target area planning data is generated, including:

[0031] According to the target position sequence in the motion prediction data, the area range of the target object activity is calculated to generate preliminary area division data;

[0032] According to the static background characteristic data, the preliminary area division data is constrained and adjusted through the scene boundary information, the invalid areas that cannot cover the target behavior are eliminated, and the optimized area data is generated;

[0033] According to the optimized regional data, the region is divided into sub-regions with different priorities to generate hierarchical regional data; the priority of the sub-region is divided by calculating the priority score and using the preset threshold. The formula for the priority score is as follows:

[0034] ,

[0035] in, For the Priority scoring for each sub-region, For the target object, The time accumulation value of each sub-area, , and is the weight coefficient, is the total number of scene features, For the scene The weight coefficient of each scene feature, For the Sub-regions The response value of the scene feature, , For the The center position vector of each sub-region, For the The position vector of each scene feature, For the The standard deviation of the Gaussian distribution of scene features, is a natural constant, For the The distance from the sub-region center to the regional center, is the total number of dynamic environment characteristics, For the Sub-regions The response value of a dynamic environmental characteristic, , For the The position vector of the dynamic environment characteristic, For the The standard deviation of the Gaussian distribution of dynamic environmental characteristics;

[0036] Setting buffer parameters for high priority sub-regions in the hierarchical regional data to form dynamic priority regional planning data;

[0037] The dynamic priority area planning data is merged with the low priority sub-areas to generate the target area planning data.

[0038] Preferably, shooting path data is formulated according to the target area planning data to generate a preliminary shooting plan, which includes the movement trajectory, shooting angle and focal length change of the camera device, including:

[0039] According to the planning data of the target area, high-priority sub-areas are extracted in chronological order to generate regional priority paths;

[0040] According to the shooting equipment parameters and the behavior changes of the target object, the regional priority path is adjusted to generate path adjustment data;

[0041] The path adjustment data is refined into shooting sub-paths, and the shooting angle and focal length changes of each path are specified to form a preliminary shooting plan.

[0042] Preferably, the captured images are analyzed in real time to evaluate the quality of the captured images and generate image quality feedback data, including:

[0043] Analyze the continuously shot images in segments, match each segment of the image data with the features of the target object, extract the integrity index of the target object in the image, and generate the target integrity data;

[0044] According to the target integrity data, the color distribution consistency, dynamic blur level and background interference level in the picture are analyzed to generate color consistency data, dynamic clarity data and background interference data respectively;

[0045] The picture is scored based on the color consistency data, dynamic clarity data and background interference data to generate the picture quality evaluation index; the picture quality evaluation index is calculated by the following formula:

[0046] ,

[0047] in, is an indicator for image quality evaluation. is the total number of screen partitions, , and is the weight coefficient, For the In the partition The first The pixel brightness gradient, For the The total number of pixels contained in the detected objects, For the The detected object is The area occupied by each partition is For the The total number of detected objects in the partitions, For the The pixel brightness gradient modulus in each partition is and The start time and end time are respectively, For the The variance of the background pixel grayscale in each partition, For the The average grayscale value of background pixels in each partition;

[0048] According to the image quality evaluation index, the picture areas that are below the preset standard are identified, marked as image quality deviation areas, and image quality feedback data is generated.

[0049] Preferably, the shooting instructions of the step-by-step shooting instruction set are optimized according to the image quality feedback data to obtain an optimized shooting instruction set, including:

[0050] The image quality feedback data is correlated and analyzed with the current step-by-step shooting instruction set, the instruction parameters causing the image quality deviation are extracted, and the parameter deviation data is generated;

[0051] According to the parameter deviation data, the movement trajectory, shooting angle and focal length change of the corresponding device in the step-by-step shooting instruction set are adjusted to generate optimized parameter data;

[0052] According to the optimization parameter data, the priority of the shooting instructions in each time period in the step-by-step shooting instruction set is reallocated to form an optimized shooting instruction set.

[0053] In a second aspect, a tracking system based on intelligent big data is provided, the system comprising:

[0054] Data acquisition module, used to obtain real-time image data of the target object, scene environment parameter data and auxiliary behavior data;

[0055] The image data processing module is used to perform frame processing on the real-time image data, extract the frames with significant changes in the picture, and generate key frame data;

[0056] The environmental data processing module is used to classify and analyze the scene environmental parameter data and extract dynamic environmental characteristic data and static background characteristic data;

[0057] A behavior data processing module is used to analyze the auxiliary behavior data, extract the movement pattern of the target object, and generate behavior characteristic data;

[0058] A motion prediction module is used to calculate the preliminary position and motion trend of the target object based on the key frame data and the behavior characteristic data in combination with the dynamic environment characteristics, and generate motion prediction data;

[0059] The area planning module is used to divide the activity range of the target object according to the motion prediction data and the static background characteristics, and generate the target area planning data;

[0060] A shooting plan generation module is used to formulate shooting path data according to target area planning data and generate a preliminary shooting plan, which includes the movement trajectory of the camera device, shooting angle and focal length change;

[0061] A shooting plan refinement module is used to refine the preliminary shooting plan into a step-by-step shooting instruction set within a time period, and dynamically adjust the shooting instructions of the step-by-step shooting instruction set according to the real-time changes in the behavior of the target object;

[0062] A shooting instruction execution module is used to execute the step-by-step shooting instruction set in real time and control the device to obtain continuous shooting images;

[0063] The image quality feedback module is used to analyze the captured images in real time, evaluate the quality of the captured images, and generate image quality feedback data;

[0064] A shooting instruction optimization module, used for optimizing the shooting instructions of the step-by-step shooting instruction set according to the image quality feedback data;

[0065] The storage module is used to store the continuously shot images and related dynamic data as an expandable multi-dimensional data set.

[0066] The above solution of the present invention includes at least the following beneficial effects:

[0067] First, by processing the real-time image data in frames, key frame data with significant changes in the picture are extracted, solving the problem of inaccurate picture capture caused by fixed algorithms and single data sources in traditional methods. The generation of key frame data can ensure the rapid capture of the main features of the target object in complex motion scenes. In addition, by classifying and analyzing the scene environment parameter data, the dynamic environment characteristics and static background characteristics are extracted, realizing the system's comprehensive perception of dynamic environmental changes. Especially in scenes such as football games, this feature can dynamically adjust the focus on multiple fast-moving targets, such as players, footballs, and referees, avoiding misjudgments and target losses caused by a single rule-driven algorithm.

[0068] Secondly, in view of the problem that the existing technology is difficult to predict the target movement trend, the present invention calculates the movement trend of the target object and generates motion prediction data by combining key frame data, dynamic environment characteristics and behavioral characteristics of the target object. The motion prediction data generates a multi-time period target position sequence by eliminating abnormal paths, which not only improves the accuracy of motion prediction, but also ensures the real-time response capability to the target object in complex scenes. For example, in a football match, this method can simultaneously predict the movement trend of the players and the trajectory changes of the football, thereby effectively avoiding the omission of key scenes.

[0069] Next, the present invention divides the activity range of the target object according to the motion prediction data and static background characteristics, and generates target area planning data. The area planning data ensures the system's accurate capture of high-priority areas by dynamically adjusting the area priority, significantly improving the efficiency and flexibility of shooting path planning. Compared with the traditional fixed-track camera system, the present invention can quickly adjust the path in a real-time environment to adapt to the rapid movement of the target object and scene changes, and is particularly suitable for live broadcasting of events with complex venues and diverse target distribution.

[0070] Finally, the present invention evaluates the picture quality in real time by generating picture quality feedback data during the shooting process. By integrating multi-dimensional picture quality indicators such as target integrity, dynamic clarity, and background interference, this method can timely mark low-quality picture areas, and dynamically adjust the movement trajectory, shooting angle, and focal length of the device by optimizing the shooting instruction set, forming a closed-loop optimization mechanism. This method ensures the clarity and continuity of the captured picture, and effectively solves the technical defects of the traditional system that the live broadcast effect is reduced due to picture quality problems.

[0071] In summary, the present invention has significant advantages in continuous tracking of target objects and optimization of picture quality in complex dynamic environments. It not only improves the intelligence level of the system, but also significantly enhances the robustness and adaptability of the system in real-time scenarios, especially showing good application prospects in live broadcasts of sports events and other high-dynamic scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 It is a flowchart of a tracking method based on intelligent big data provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0073] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0074] like Figure 1 As shown, an embodiment of the present invention proposes a tracking method based on smart big data, and the method includes:

[0075] S100, acquiring real-time image data, scene environment parameter data, and auxiliary behavior data of a target object;

[0076] S200, performing frame processing on the real-time image data, extracting frames with significant changes in the picture, and generating key frame data;

[0077] S300, classify and analyze the scene environment parameter data to extract dynamic environment characteristic data and static background characteristic data;

[0078] S400, parsing the auxiliary behavior data, extracting the movement pattern of the target object, and generating behavior characteristic data;

[0079] S500, calculating the preliminary position and motion trend of the target object according to the key frame data and the behavior characteristic data in combination with the dynamic environment characteristics, and generating motion prediction data;

[0080] S600, dividing the activity range of the target object according to the motion prediction data and the static background characteristics, and generating target area planning data;

[0081] S700, formulating shooting path data according to the target area planning data, and generating a preliminary shooting plan, wherein the preliminary shooting plan includes a moving track of the camera device, a shooting angle, and a focal length change;

[0082] S800, breaking down the preliminary shooting plan into a step-by-step shooting instruction set within a time period, and dynamically adjusting the shooting instructions of the step-by-step shooting instruction set according to the real-time changes in the behavior of the target object;

[0083] S900, executing the step-by-step shooting instruction set in real time, and controlling the device to obtain continuous shooting images;

[0084] S1000, analyzing the captured images in real time, evaluating the quality of the captured images, and generating image quality feedback data;

[0085] S1100, optimizing the shooting instructions of the step-by-step shooting instruction set according to the image quality feedback data to obtain an optimized shooting instruction set;

[0086] S1200: Store the continuously shot images and related dynamic data as an expandable multi-dimensional dataset.

[0087] In the embodiment of the present invention, firstly, the frame processing of real-time image data can extract the key frames of the target object, and can capture the significant features and change trends of the target object, which is particularly critical in fast-motion scenes. At the same time, the classification analysis of environmental parameter data generates dynamic environmental characteristics and static background characteristics, which can perceive the changes in complex scenes, and extract the motion pattern of the target object in combination with auxiliary behavior data analysis. With the support of behavioral characteristic data, a motion trend model of the target object can be established, thereby laying the foundation for subsequent prediction and planning.

[0088] In the process of generating motion prediction data, the initial position distribution of the target object and the movement trend within the time period are calculated by combining key frame data, dynamic environment characteristics and behavioral characteristics, thereby optimizing the accuracy of the target activity trajectory. In this process, abnormal paths are eliminated and target position sequences in multiple time periods are generated, making the prediction of target motion more accurate and reliable. By combining motion prediction data with static background characteristic data, the activity range of the target object can be delineated and target area planning data can be generated. Regional division takes into account environmental constraints and priorities, providing high-quality input for subsequent path planning.

[0089] Based on the target area planning data, the system develops a shooting path, including the device movement trajectory, shooting angle adjustment, and focal length change. In this process, the coherence and stability of the path planning are optimized by combining dynamic environment data and behavioral characteristic data. By refining the shooting path by time segmentation, the system can adjust the shooting strategy in real time to ensure continuous tracking of the target object. Finally, the generation of image quality feedback data comprehensively evaluates the quality of the captured image through multi-dimensional indicators, such as target integrity, color consistency, and background interference. These feedback data are directly used to optimize the shooting instructions, enabling the system to adjust the device parameters in real time and form a closed-loop optimization mechanism, thereby achieving high-quality tracking of the target object. This end-to-end efficient process significantly improves the shooting effect in complex dynamic scenes.

[0090] More specifically, in performing frame processing on the real-time image data, extracting frames with significant changes in the picture, and generating key frame data, the specific steps include:

[0091] Adjacent frames are differentiated to identify areas with large changes in pixel values.

[0092] Apply threshold filtering to keep frames with significant pixel differences and remove frames with small changes.

[0093] The filtered frames are further analyzed and the target detection algorithm is used to confirm whether they contain important information about the target object, such as changes in movement or position.

[0094] The key frame data finally generated is a subset extracted from the video stream, containing the moments of significant changes of the target object, which is used for subsequent motion analysis and prediction.

[0095] More specifically, in classifying and analyzing the scene environment parameter data and extracting the dynamic environment characteristic data and the static background characteristic data, the specific steps include:

[0096] Scene environment parameter data refers to external factors that affect target motion and image capture, including dynamic and static information. For example, dynamic environment parameters can be light source position, wind speed change, etc., and static background parameters can be fixed site boundaries or obstacle distribution.

[0097] Classification analysis is to divide the environmental parameters in the scene into two categories: dynamic and static:

[0098] Dynamic environmental characteristic data describes environmental characteristics that change over time, such as changes in lighting and the location of moving obstacles.

[0099] Static background feature data describes constant or slowly changing scene features, such as fixed buildings or site boundaries.

[0100] The main steps of data classification are:

[0101] Obtain scene environment parameters through sensors or external data sources.

[0102] Analyze the time variation trend of each parameter to determine whether it is dynamic or static.

[0103] Dynamic environment characteristics are further refined into key factors that affect target motion, such as the brightness and direction changes of dynamic light sources.

[0104] Static background characteristics are mainly spatial distribution, including the boundaries and distribution of fixed obstacles.

[0105] More specifically, in breaking down the preliminary shooting plan into a step-by-step shooting instruction set within a time period, and dynamically adjusting the shooting instructions of the step-by-step shooting instruction set according to the real-time changes in the behavior of the target object, the specific steps include:

[0106] The preliminary shooting plan is a rough shooting path generated based on the target's motion trend and area priority.

[0107] Refining the plan into a step-by-step shooting instruction set means dividing the preliminary plan into multiple specific shooting steps according to time periods, each step including the movement trajectory of the device, shooting angle and focal length parameters.

[0108] The dynamic adjustment of the step-by-step instruction set depends on the real-time behavior changes of the target object. For example:

[0109] If the target object changes its direction of movement, the shooting angle of the device needs to be adjusted.

[0110] If the target accelerates, you may need to shorten the shooting interval to improve the continuous shooting capability.

[0111] When a target enters a low-priority area, the system can temporarily reduce its attention to that area and respond to high-priority areas first.

[0112] More specifically, according to the step-by-step shooting instruction set, it is executed in real time, and the device is controlled to obtain continuous shooting images:

[0113] Real-time execution refers to sending control signals to the camera device according to the step-by-step shooting instruction set, so that it can complete operations such as trajectory movement, shooting angle adjustment and focal length change. The goal is to ensure that the device captures the target object continuously and without omission, especially when the target moves quickly or the environment changes complexly. The realization of the captured picture requires closed-loop control that combines shooting instructions with real-time feedback. For example: the camera adjusts the angle and focal length according to the instructions to ensure that the target object is always in the center of the picture. When the real-time feedback detects that the target deviates or the image quality decreases, the system can immediately adjust the shooting parameters.

[0114] In a preferred embodiment of the present invention, the auxiliary behavior data is parsed, the movement pattern of the target object is extracted, and the behavior characteristic data is generated, including:

[0115] Analyze auxiliary behavior data, identify the behavior frequency and change range of the target object, and extract behavior characteristics;

[0116] According to the time sequence information of the target object, the movement direction, speed and trajectory form in the behavior characteristics are analyzed to generate the movement pattern of the target object;

[0117] Combine the target object's movement pattern and behavioral characteristics to form behavioral characteristic data.

[0118] In the embodiment of the present invention, the motion pattern of the target object is extracted by analyzing the auxiliary behavior data, thereby enhancing the system's ability to perceive the dynamic changes of the target. The analysis of the behavior data mainly includes the analysis of the target action frequency, change amplitude and timing information, generating a complete feature of the target behavior. These behavior characteristic data show extremely high accuracy in describing the movement direction, speed and trajectory morphology of the target object, especially in scenes with complex and diverse target behaviors. This feature is particularly important.

[0119] In addition, the generation of behavioral characteristic data enables the system to better understand the movement patterns of the target object by establishing a dynamic behavior model of the target object. In dynamic scenes, the behavior of the target object may be affected by environmental changes. The accurate extraction of behavioral characteristic data provides a reliable basis for the system to predict the movement trend of the target object. Through such feature extraction, the system can analyze the movement trajectory of the target object more quickly and accurately, providing support for subsequent path planning and optimization. This method significantly improves the robustness and accuracy of target tracking in complex dynamic environments.

[0120] More specifically, in parsing the auxiliary behavior data, identifying the behavior frequency and change range of the target object, and extracting the behavior features, the specific steps include:

[0121] Identification of behavior frequency: Use time series analysis methods to collect statistics on auxiliary behavior data and extract the frequency of repeated occurrence of a certain behavior. For example, decompose the behavior signal into a spectrum through Fourier transform to identify the main cycle of the behavior.

[0122] Analysis of the amplitude of change: By comparing the changes in behavioral parameters such as movement amplitude and displacement range in different time periods, the amplitude of fluctuation is calculated. It can be described by standard deviation or rate of change formula.

[0123] Extraction of behavioral features: Input frequency and amplitude data into the feature extraction model to generate feature vectors that represent the key behavioral patterns of the target object, such as gait, arm swing, etc.

[0124] More specifically, in analyzing the movement direction, speed and trajectory form in the behavior characteristics according to the time sequence information of the target object, and generating the movement pattern of the target object, the specific steps include:

[0125] Analysis of motion direction: Based on the position and direction information in the auxiliary behavior data, such as coordinate changes or direction vectors, the motion vector is calculated using the time difference method.

[0126] Speed ​​calculation: The instantaneous speed of the target object is obtained through the modulus of the motion direction vector.

[0127] Trajectory morphology construction: The movement direction and speed at multiple time points are combined to generate the trajectory of the target object. Bezier curves or Kalman filtering algorithms can be used to smooth the trajectory to form a continuous movement pattern.

[0128] Generation of motion patterns: Based on the above analysis, machine learning models such as cluster analysis are used to classify different motion patterns into static, linear motion, or complex path motion.

[0129] More specifically, in combining the movement pattern and behavior characteristics of the target object to form behavior characteristic data, the specific steps include:

[0130] Feature fusion: Combine the extracted behavioral features, frequency, amplitude and motion patterns, such as direction, speed and trajectory, and use feature concatenation or dimensionality reduction methods, such as principal component analysis, to generate a fused feature vector.

[0131] Generation of behavioral characteristics: Based on the fused features, a behavioral characteristic dataset is constructed, including key descriptions of behavioral patterns, such as activity types with high repetition frequency and path morphology of complex movements.

[0132] Behavior modeling: Through behavioral characteristic data, prediction models such as long short-term memory networks (LSTM) can be used to predict the next behavior of the target object, forming real-time feedback or for decision support.

[0133] In a preferred embodiment of the present invention, the preliminary position and motion trend of the target object are calculated based on the key frame data and the behavior characteristic data, combined with the dynamic environment characteristics, to generate motion prediction data, including:

[0134] Synchronously analyze the key frame data and the behavior characteristic data, and generate preliminary position distribution data by combining the position distribution and motion pattern of the target object;

[0135] According to the preliminary position distribution data and the dynamic environment characteristics in the environment characteristic data, the motion vector of the target object in different time periods is calculated to generate motion trend data; the motion vector is calculated by the following formula;

[0136] ,

[0137] in, For the target object at time The motion vector of For the target object at time point The motion vector of is the time step, For in time The total number of dynamic environments at the target object's location, For in time The number of target object behavior characteristics, For the The weight of the dynamic environment characteristics, For the The weight of the behavioral characteristics, For the The impact vector of dynamic environment characteristics on the target object, , For the target object at time The position vector of For the The position vector of the dynamic environment characteristic, For the The influence vector of the behavior characteristics on the target object, , For the The position vector of the behavioral characteristics, and is the magnitude of the corresponding vector;

[0138] Perform correlation analysis on the preliminary location distribution data and motion trend data, eliminate abnormal motion paths that are inconsistent with the dynamic environment, and generate location prediction data;

[0139] The position prediction data is divided into time periods to form a target position sequence of multiple time periods and generate motion prediction data.

[0140] In the embodiment of the present invention, the movement trend of the target object in the time period is generated by combining the key frame data and the behavior characteristic data, providing a highly accurate input for the target motion prediction. By synchronously analyzing the key frame data of the target object, the system can quickly capture the position and change characteristics of the target in the picture, while the behavior characteristic data provides accurate modeling for the dynamic laws of the target object. The system uses the influence of dynamic environmental characteristics on the target motion, combined with the trajectory information of the target object in the time period, to generate preliminary position distribution data.

[0141] On this basis, the system removes abnormal data in the target motion path to ensure more accurate prediction of motion trends. By dividing into multiple time periods, the system generates multiple target position sequences, making the motion prediction results more complete in the time dimension. Combined with the changing laws of the dynamic environment, the prediction of motion trends significantly improves the stability of target tracking. This method is particularly suitable for complex dynamic environments, such as multi-person scenes or fast target motion scenes, and provides reliable support for subsequent area division and path planning.

[0142] More specifically, the key frame data and the behavior characteristic data are synchronously analyzed, and the position distribution and motion pattern of the target object are combined to generate preliminary position distribution data. The specific steps include:

[0143] Synchronous analysis of key frame data and behavioral characteristic data: Extract significantly changed frames from the real-time image of the target object, indicating the key moments of movement or behavior. Each key frame contains a timestamp, spatial coordinates, and the characteristics of the target object in the frame. Including the target object's behavior frequency, amplitude, direction, speed, and trajectory information, representing the object's movement pattern over a period of time. Align the key frame data and behavioral characteristic data through timestamps to ensure the consistency of the two types of data in the time dimension. Linear interpolation methods can be used to align and supplement data that do not completely match in time.

[0144] Generate location distribution data: Combine the spatial coordinates in the keyframes and the description of the motion pattern in the behavior characteristic data to cluster the historical locations of the target object and obtain the main activity area of ​​the target object. Use the motion direction and speed data to dynamically expand the range of the location distribution. For example, calculate the possible movement radius of the target object based on the speed. The preliminary location distribution data is expressed in the form of spatial probability density, showing the possibility of the target object appearing in different areas.

[0145] More specifically, the preliminary location distribution data and the motion trend data are correlated and analyzed, abnormal motion paths that are inconsistent with the dynamic environment are eliminated, and location prediction data is generated. The specific steps include:

[0146] Correlation analysis between preliminary position distribution data and motion trend data: The motion vector of the target object, including direction and speed, calculated by analyzing dynamic environmental characteristics, such as wind speed and obstacle changes, describes the movement direction and trend of the target object in the future. The preliminary position distribution data is matched with the motion trend data to analyze the rationality of the target object's movement at different locations. For example, some motion paths may be considered abnormal because they conflict with the position of obstacles.

[0147] Eliminate abnormal motion paths: Use dynamic environment characteristics, such as obstacle locations and dynamic changes, to verify the rationality of motion paths. For example, if the target object's motion trend points to a known inaccessible area, the path is marked as abnormal. Based on the threshold, paths below the preset threshold will be filtered out, and reasonable paths will be retained. The position prediction data is the probability distribution of the possible location of the target object after eliminating abnormal paths.

[0148] More specifically, for dividing the position prediction data into time periods, forming a target position sequence of multiple time periods, and generating motion prediction data, the specific steps include:

[0149] Time segmentation: The location prediction data is divided into multiple segments according to the time dimension, and the possible location of the target object in each segment is represented by a location sequence.

[0150] Time period length determination: The length of the time period is determined based on the motion trend and behavior characteristics. For example, more stable behaviors, such as uniform motion, can use longer time periods, while highly dynamic behaviors require shorter time periods to improve prediction accuracy.

[0151] Target position sequence generation: Position sequence representation: In each time period, the possible positions of the target object are represented by a series of spatial coordinate points.

[0152] Trajectory optimization: Through trajectory smoothing algorithms such as Kalman filtering, the position sequence is optimized to ensure the continuity between sequences.

[0153] Generate motion prediction data: Integrate the target position sequences of all time periods to form the motion prediction data of the target object. These data not only contain the possible positions of the target object in each future time period, but also reflect the overall motion trend.

[0154] Output form: Motion prediction data can be expressed in the form of time series and spatial distribution, providing an accurate basis for device path planning.

[0155] In a preferred embodiment of the present invention, the activity range of the target object is divided according to the motion prediction data and the static background characteristics, and the target area planning data is generated, including:

[0156] According to the target position sequence in the motion prediction data, the area range of the target object activity is calculated to generate preliminary area division data;

[0157] According to the static background characteristic data, the preliminary area division data is constrained and adjusted through the scene boundary information, the invalid areas that cannot cover the target behavior are eliminated, and the optimized area data is generated;

[0158] According to the optimized regional data, the region is divided into sub-regions with different priorities to generate hierarchical regional data; the priority of the sub-region is divided by calculating the priority score and using the preset threshold. The formula for the priority score is as follows:

[0159] ,

[0160] in, For the Priority scoring for each sub-region, For the target object, The time accumulation value of each sub-area, , and is the weight coefficient, is the total number of scene features, For the scene The weight coefficient of each scene feature, For the Sub-regions The response value of the scene feature, , For the The center position vector of each sub-region, For the The position vector of each scene feature, For the The standard deviation of the Gaussian distribution of scene features, is a natural constant, For the The distance from the sub-region center to the regional center, is the total number of dynamic environment characteristics, For the Sub-regions The response value of a dynamic environmental characteristic, , For the The position vector of the dynamic environment characteristic, For the The standard deviation of the Gaussian distribution of dynamic environmental characteristics;

[0161] Setting buffer parameters for high priority sub-regions in the hierarchical regional data to form dynamic priority regional planning data;

[0162] The dynamic priority area planning data is merged with the low priority sub-areas to generate the target area planning data.

[0163] In the embodiment of the present invention, the division of the target activity range combines motion prediction data and static background characteristic data, so that the system can generate highly accurate target area planning data. By analyzing the movement trend of the target within the time period, combined with the scene boundary information and dynamic environment constraints, the system accurately divides the target activity range. The area division takes into account the behavior priority of the target object and the influence of the dynamic environment, so that the area planning results are more in line with the actual scene requirements.

[0164] By calculating the priority of each sub-area, the system can provide a basis for shooting path planning. The priority calculation process combines the response value of the sub-area to the scene characteristics and the response value of the dynamic environment, making the area division highly flexible and adaptable. The Gaussian distribution calculation method of the area priority ensures that the weight of the area far away from the target object gradually decreases, while the area close to the target object obtains a higher priority, thereby improving the rationality of the target area planning and the accuracy of the shooting path. This method effectively improves the efficiency of area division in dynamic scenes.

[0165] More specifically, buffer parameters are set for high-priority sub-regions in the hierarchical region data to form dynamic priority region planning data. The specific steps include:

[0166] Setting of buffer parameters: To prevent the area from exceeding the planned scope due to changes in the behavior of the target object, a buffer range is set around each high-priority sub-area. Buffer parameters can include: forming a buffer zone by increasing the sub-area boundary range; adding a certain margin for the time the target object may stay in the area. Buffer parameters are related to the target object's movement speed and direction change frequency; for example, when the target object moves faster, the buffer range should be increased to cover the possible extension position.

[0167] Forming dynamic priority area planning data: Combined with buffer parameters, high priority sub-areas are adjusted in real time to form dynamic priority areas to ensure that the behavior changes of target objects in the area do not exceed the planning range. Through dynamic priority area planning data, changes in target behavior can be better coordinated with the following capabilities of the shooting equipment, improving real-time responsiveness.

[0168] In a preferred embodiment of the present invention, shooting path data is formulated according to the target area planning data to generate a preliminary shooting plan, which includes the moving track, shooting angle and focal length change of the camera device, including:

[0169] According to the planning data of the target area, high-priority sub-areas are extracted in chronological order to generate regional priority paths;

[0170] According to the shooting equipment parameters and the behavior changes of the target object, the regional priority path is adjusted to generate path adjustment data;

[0171] The path adjustment data is refined into shooting sub-paths, and the shooting angle and focal length changes of each path are specified to form a preliminary shooting plan.

[0172] In the embodiment of the present invention, through the in-depth analysis of the target area planning data, the system can generate a shooting path with strong coherence to ensure the continuous tracking of the target object by the device. The path planning is combined with the division of target priority areas within the time period. Especially in dynamic scenes, by optimizing the time sequence of high priority areas, the system can ensure that the device always accurately captures the target.

[0173] In addition, by dynamically adjusting the path parameters, the shooting path planning has the ability to adapt in real time. The system's reasonable allocation of shooting angles and focal length changes makes the shooting results more consistent with the movement pattern and trajectory of the target object. This solution significantly improves the efficiency of tracking shooting in complex scenes, while reducing the problem of image loss caused by improper path adjustment of the device.

[0174] More specifically, according to the planning data of the target area, high-priority sub-areas are extracted in chronological order to generate regional priority paths. The specific steps include:

[0175] Extract high-priority sub-areas: Combine the time information in the target area planning data to extract high-priority sub-areas where the target object appears first in different time periods. Sort by priority score in the sub-area and the time sequence information of the target object; sub-areas with earlier timestamps are included in the path planning first.

[0176] Generate regional priority paths: Take the center point of the high-priority sub-region as the path node, and generate the initial path based on the spatial position and time sequence. To ensure the continuity and rationality of the path, the traveling salesman problem TSP optimization algorithm can be used to calculate the shortest path to reduce the time consumption and energy waste during the movement of the shooting equipment.

[0177] More specifically, according to the shooting device parameters and the behavior changes of the target object, the regional priority path is adjusted to generate the path adjustment data. The specific steps include:

[0178] Consideration of shooting equipment parameters: such as the moving speed of the equipment, shooting angle range, lens focus adjustment speed, etc. will affect the path adjustment. The behavior of the target object (such as sudden turn or stop) requires real-time adjustment of the path to ensure that the device can keep shooting the target object.

[0179] Dynamic optimization of path adjustment: According to the behavior changes of the target object, combined with the real-time feedback of the shooting device, the regional priority path is adjusted. For example, if the target object does not appear at the current path node, the device waiting time is extended or the path is replanned. If the target object suddenly changes direction, the path needs to be recalculated to adjust the movement direction of the device. The adjusted path data includes the coordinates of each step of the device's movement, the movement time, and the stop position.

[0180] More specifically, the path adjustment data is refined into shooting sub-paths, and the shooting angle and focal length changes of each path are specified to form a preliminary shooting plan. The specific steps include:

[0181] Refine the shooting sub-path: According to the adjusted path, the entire path is subdivided into multiple sub-paths, and each sub-path corresponds to a shooting task. In the subdivision process, the behavioral characteristics of the target object and the performance of the equipment are considered. For example, in the path section where the target object accelerates, the shooting device needs to move to the estimated position in advance; in the complex behavior area, the shooting path should shorten the step length to obtain more details.

[0182] Specify shooting angle and focal length change: Determine the shooting angle range of the device according to the target object's movement direction and position in each sub-path. Focal length change control: Combined with the real-time distance between the target object and the device, dynamically adjust the focal length to keep the target object clearly presented.

[0183] Formation of a preliminary shooting plan: The preliminary shooting plan includes specific instructions such as the equipment movement trajectory of the sub-path, the shooting angle range of each path, and the focus adjustment requirements.

[0184] In a preferred embodiment of the present invention, the captured images are analyzed in real time, the quality of the captured images is evaluated, and image quality feedback data is generated, including:

[0185] Analyze the continuously shot images in segments, match each segment of the image data with the features of the target object, extract the integrity index of the target object in the image, and generate target integrity data;

[0186] According to the target integrity data, the color distribution consistency, dynamic blur level and background interference level in the picture are analyzed to generate color consistency data, dynamic clarity data and background interference data respectively;

[0187] The picture is scored based on the color consistency data, dynamic clarity data and background interference data to generate the picture quality evaluation index; the picture quality evaluation index is calculated by the following formula:

[0188] ,

[0189] in, is an indicator for image quality evaluation. is the total number of screen partitions, , and is the weight coefficient, For the In the partition The first The pixel brightness gradient, For the The total number of pixels contained in the detected objects, For the The detected object is The area occupied by each partition is For the The total number of detected objects in the partitions, For the The pixel brightness gradient modulus in each partition is and The start time and end time are respectively, For the The variance of the background pixel grayscale in each partition, For the The average grayscale value of background pixels in each partition;

[0190] According to the image quality evaluation index, the picture areas that are below the preset standard are identified, marked as image quality deviation areas, and image quality feedback data is generated.

[0191] In the embodiment of the present invention, through the generation of image quality feedback data, the system realizes a comprehensive evaluation of the captured images, significantly improving the efficiency of image quality optimization. The image quality evaluation integrates indicators of multiple dimensions, including target integrity, dynamic clarity, and background interference, and quantitatively analyzes the image quality through scoring. The combination of these indicators enables the system to quickly mark the image areas that are below the standard and provides a clear basis for subsequent optimization.

[0192] The system's image quality assessment is not only applicable to static target scenes, but also shows high adaptability in dynamic scenes. Through dynamic monitoring of image quality, the system can identify image quality degradation caused by environmental changes or improper equipment parameters, and adjust the shooting strategy in time through a feedback mechanism. This process significantly reduces the impact of image quality degradation and ensures that the shooting results always meet high quality standards.

[0193] More specifically, the continuously shot images are analyzed in segments, each segment of the image data is matched with the features of the target object, the integrity index of the target object in the image is extracted, and the target integrity data is generated. The specific steps include:

[0194] Segment parsing: The continuously shot video stream is segmented by time or frame number, and each segment can contain tens to hundreds of frames. Segmentation can be based on time intervals, such as one segment per second, or key frame extraction technology, such as generating a segment every time the position of the target object changes significantly.

[0195] Feature matching: Obtain the features of the target object from auxiliary behavior data or real-time keyframes, such as outline, color, motion pattern, etc. Use feature matching algorithms, such as template matching or deep learning-based target detection models, to compare the target in the picture with known features and identify the boundary, shape, and position of the target.

[0196] Integrity index extraction: The pixel ratio of the target object in the picture is used to determine whether the target is completely captured. The indicators are recorded as a time series to form target integrity data.

[0197] More specifically, according to the target integrity data, the color distribution consistency, dynamic blur level and background interference level in the picture are analyzed to generate color consistency data, dynamic clarity data and background interference data respectively. The specific steps include:

[0198] Color distribution consistency analysis: Calculate the color characteristics of the target object in the entire picture, such as RGB distribution and saturation. Use histogram similarity or color clustering technology to evaluate the color differences between different pictures. Generate color consistency data, expressed as distribution variance or histogram matching score.

[0199] Motion blur analysis: Use Laplace transform or gradient detection to evaluate image clarity. Perform blur comparison on key frames in the motion trajectory to analyze the blur level during dynamic shooting. The blur level is indicated by the gradient change amplitude, and a low gradient amplitude indicates blurry images.

[0200] Background interference level analysis: The target object and background area are separated by the target detection algorithm, and the factors in the background that may interfere with the clarity of the target, such as dynamic objects or complex textures, are counted. The background interference level data represents the pixel ratio and change amplitude of the interference in the background area.

[0201] In a preferred embodiment of the present invention, the shooting instructions of the step-by-step shooting instruction set are optimized according to the image quality feedback data to obtain the optimized shooting instruction set, including:

[0202] The image quality feedback data is correlated and analyzed with the current step-by-step shooting instruction set, the instruction parameters causing the image quality deviation are extracted, and the parameter deviation data is generated;

[0203] According to the parameter deviation data, the movement trajectory, shooting angle and focal length change of the corresponding device in the step-by-step shooting instruction set are adjusted to generate optimized parameter data;

[0204] According to the optimization parameter data, the priority of the shooting instructions in each time period in the step-by-step shooting instruction set is reallocated to form an optimized shooting instruction set.

[0205] In the embodiment of the present invention, by optimizing the shooting instruction set, the system builds a real-time closed-loop optimization mechanism to make the shooting path and parameter adjustment of the device more efficient. The correlation analysis of the image quality feedback data and the current shooting instruction enables the system to quickly identify the main problems affecting the image quality and dynamically adjust the motion trajectory, shooting angle and focal length parameters of the device.

[0206] The optimized shooting instructions can adapt to the complexity of target behavior and environmental changes in real time, especially in multi-target scenes, ensuring the continuous tracking of key target objects by adjusting priorities. This mechanism significantly improves the flexibility and image quality stability of shooting in complex scenes, providing basic support for subsequent high-quality image analysis.

[0207] More specifically, the image quality feedback data is correlated with the current step-by-step shooting instruction set for analysis, the instruction parameters that cause the image quality deviation are extracted, and the parameter deviation data is generated. The specific steps include:

[0208] Correlation analysis of image quality feedback data and instruction set: Image quality feedback data, such as color consistency, blur level, background interference, and parameters in the shooting instruction set, such as device movement trajectory, shooting angle, focus adjustment, etc., are correlated. If the target integrity data is low, analyze whether the shooting angle is not correctly aimed at the target. If the blur level is high, detect whether the device movement speed and focus adjustment are too fast. If the color consistency is low, evaluate whether the ambient lighting parameters match.

[0209] Generate parameter deviation data: The deviation data is expressed as specific shooting parameters that need to be adjusted, such as moving speed, shooting angle error range, focal length correction value, etc.

[0210] More specifically, according to the optimization parameter data, the priority of the shooting instructions in each time period in the step-by-step shooting instruction set is reallocated to form an optimized shooting instruction set. The specific steps include:

[0211] Reassignment of command priorities: Reassign shooting tasks in different time periods based on the dynamic changes of the target object and the equipment adjustment capabilities. Time periods in high-dynamic behavior areas and key target behaviors have higher priority. Combined with target behavior prediction and optimization parameters, dynamically adjust the shooting command priority.

[0212] Optimize the formation of shooting instruction set: The optimized shooting instruction set includes the corrected device movement trajectory, shooting angle adjustment value and focal length change range. Ensure that the shooting device can capture the target object with high quality in different scenes, reduce image blur, color distortion and background interference.

[0213] An embodiment of the present invention further provides a tracking system based on intelligent big data, the system comprising:

[0214] Data acquisition module, used to obtain real-time image data of the target object, scene environment parameter data and auxiliary behavior data;

[0215] The image data processing module is used to perform frame processing on the real-time image data, extract the frames with significant changes in the picture, and generate key frame data;

[0216] The environmental data processing module is used to classify and analyze the scene environmental parameter data and extract dynamic environmental characteristic data and static background characteristic data;

[0217] A behavior data processing module is used to analyze the auxiliary behavior data, extract the movement pattern of the target object, and generate behavior characteristic data;

[0218] A motion prediction module is used to calculate the preliminary position and motion trend of the target object based on the key frame data and the behavior characteristic data in combination with the dynamic environment characteristics, and generate motion prediction data;

[0219] The area planning module is used to divide the activity range of the target object according to the motion prediction data and the static background characteristics, and generate the target area planning data;

[0220] The shooting plan generation module is used to formulate shooting path data according to the target area planning data and generate a preliminary shooting plan, which includes the movement trajectory of the camera equipment, the shooting angle and the focal length change;

[0221] A shooting plan refinement module is used to refine the preliminary shooting plan into a step-by-step shooting instruction set within a time period, and dynamically adjust the shooting instructions of the step-by-step shooting instruction set according to the real-time changes in the behavior of the target object;

[0222] A shooting instruction execution module is used to execute the step-by-step shooting instruction set in real time and control the device to obtain continuous shooting images;

[0223] The image quality feedback module is used to analyze the captured images in real time, evaluate the quality of the captured images, and generate image quality feedback data;

[0224] A shooting instruction optimization module, used for optimizing the shooting instructions of the step-by-step shooting instruction set according to the image quality feedback data;

[0225] The storage module is used to store the continuously shot images and related dynamic data as an expandable multi-dimensional data set.

[0226] It should be noted that the system is a system corresponding to the above method, and all implementation methods in the above method embodiment are applicable to this embodiment and can achieve the same technical effect.

[0227] The embodiment of the present invention further provides a computing device, including: a processor, a memory storing a computer program, and when the computer program is executed by the processor, the method described above is executed. All implementations in the above method embodiment are applicable to this embodiment and can achieve the same technical effect.

[0228] The embodiment of the present invention further provides a computer-readable storage medium storing instructions, which, when executed on a computer, enable the computer to execute the above method. All implementations in the above method embodiment are applicable to this embodiment and can achieve the same technical effect.

[0229] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A tracking method based on intelligent big data, characterized in that: The method comprises: Obtain real-time image data, scene environment parameter data, and auxiliary behavior data of the target object; Perform frame processing on real-time image data, extract frames with significant changes in the picture, and generate key frame data; Classify and analyze scene environment parameter data to extract dynamic environment characteristic data and static background characteristic data; Analyze the auxiliary behavior data, extract the movement pattern of the target object, and generate behavior characteristic data; According to the key frame data and behavior characteristic data, combined with the dynamic environment characteristics, the preliminary position and motion trend of the target object are calculated to generate motion prediction data; According to the motion prediction data and static background characteristics, the activity range of the target object is divided and the target area planning data is generated; According to the target area planning data, the shooting path data is formulated to generate a preliminary shooting plan, which includes the movement trajectory of the camera equipment, the shooting angle and the focal length change; Refine the preliminary shooting plan into a step-by-step shooting instruction set within a time period, and dynamically adjust the shooting instructions of the step-by-step shooting instruction set according to the real-time changes in the behavior of the target object; According to the step-by-step shooting instruction set, it is executed in real time, and the device is controlled to obtain continuous shooting images; Real-time analysis of the captured images, evaluation of their quality, and generation of image quality feedback data; Optimizing the shooting instructions of the step-by-step shooting instruction set according to the image quality feedback data to obtain an optimized shooting instruction set; Store continuous shooting images and related dynamic data as expandable multi-dimensional datasets.

2. The tracking method based on intelligent big data according to claim 1, characterized in that: Analyze the auxiliary behavior data, extract the movement pattern of the target object, and generate behavior characteristic data, including: Analyze auxiliary behavior data, identify the behavior frequency and change range of the target object, and extract behavior characteristics; According to the time sequence information of the target object, the movement direction, speed and trajectory form in the behavior characteristics are analyzed to generate the movement pattern of the target object; Combine the target object's movement pattern and behavioral characteristics to form behavioral characteristic data.

3. The method for following and shooting based on intelligent big data according to claim 2, characterized in that: Based on the key frame data and behavior characteristic data, combined with the dynamic environment characteristics, the preliminary position and motion trend of the target object are calculated to generate motion prediction data, including: Synchronously analyze the key frame data and the behavior characteristic data, and generate preliminary position distribution data by combining the position distribution and motion pattern of the target object; According to the preliminary position distribution data and the dynamic environment characteristics in the environment characteristic data, the motion vector of the target object in different time periods is calculated to generate motion trend data; the motion vector is calculated by the following formula; ， in, For the target object at time The motion vector of For the target object at time point The motion vector of is the time step, For in time The total number of dynamic environments at the target object's location, For in time The number of target object behavior characteristics, For the The weight of the dynamic environment characteristics, For the The weight of the behavioral characteristics, For the The impact vector of dynamic environment characteristics on the target object, , For the target object at time The position vector of For the The position vector of the dynamic environment characteristic, For the The influence vector of the behavior characteristics on the target object, , For the The position vector of the behavioral characteristics, and is the magnitude of the corresponding vector; Perform correlation analysis on the preliminary location distribution data and motion trend data, eliminate abnormal motion paths that are inconsistent with the dynamic environment, and generate location prediction data; The position prediction data is divided into time periods to form a target position sequence of multiple time periods and generate motion prediction data.

4. The tracking method based on intelligent big data according to claim 3 is characterized in that: According to the motion prediction data and static background characteristics, the activity range of the target object is divided and the target area planning data is generated, including: According to the target position sequence in the motion prediction data, the area range of the target object activity is calculated to generate preliminary area division data; According to the static background characteristic data, the preliminary area division data is constrained and adjusted through the scene boundary information, the invalid areas that cannot cover the target behavior are eliminated, and the optimized area data is generated; According to the optimized regional data, the region is divided into sub-regions with different priorities to generate hierarchical regional data; the priority of the sub-region is divided by calculating the priority score and using the preset threshold. The formula for the priority score is as follows: ， in, For the Priority scoring for each sub-region, For the target object, The time accumulation value of each sub-area, , and is the weight coefficient, is the total number of scene features, For the scene The weight coefficient of each scene feature, For the Sub-regions The response value of the scene feature, , For the The center position vector of each sub-region, For the The position vector of each scene feature, For the The standard deviation of the Gaussian distribution of scene features, is a natural constant, For the The distance from the sub-region center to the regional center, is the total number of dynamic environment characteristics, For the Sub-regions The response value of a dynamic environmental characteristic, , For the The position vector of the dynamic environment characteristic, For the The standard deviation of the Gaussian distribution of dynamic environmental characteristics; Setting buffer parameters for high priority sub-regions in the hierarchical regional data to form dynamic priority regional planning data; The dynamic priority area planning data is merged with the low priority sub-areas to generate the target area planning data.

5. The method for following and shooting based on intelligent big data according to claim 4, characterized in that: According to the target area planning data, the shooting path data is formulated and a preliminary shooting plan is generated. The preliminary shooting plan includes the movement trajectory of the camera equipment, the shooting angle and the focal length change, including: According to the planning data of the target area, high-priority sub-areas are extracted in chronological order to generate regional priority paths; According to the shooting equipment parameters and the behavior changes of the target object, the regional priority path is adjusted to generate path adjustment data; The path adjustment data is refined into shooting sub-paths, and the shooting angle and focal length changes of each path are specified to form a preliminary shooting plan.

6. The tracking method based on intelligent big data according to claim 5, characterized in that: Real-time analysis of the captured images, evaluation of the quality of the captured images, and generation of image quality feedback data, including: Analyze the continuously shot images in segments, match each segment of the image data with the features of the target object, extract the integrity index of the target object in the image, and generate the target integrity data; According to the target integrity data, the color distribution consistency, dynamic blur level and background interference level in the picture are analyzed to generate color consistency data, dynamic clarity data and background interference data respectively; The picture is scored based on the color consistency data, dynamic clarity data and background interference data to generate the picture quality evaluation index; the picture quality evaluation index is calculated by the following formula: ； in, is an indicator for image quality evaluation. is the total number of screen partitions, , and is the weight coefficient, For the In the partition The first The pixel brightness gradient, For the The total number of pixels contained in the detected objects, For the The detected object is The area occupied by each partition is For the The total number of detected objects in the partitions, For the The pixel brightness gradient modulus in each partition is and The start time and end time are respectively, For the The variance of the background pixel grayscale in each partition, For the The average grayscale value of background pixels in each partition; According to the image quality evaluation index, the picture areas that are below the preset standard are identified, marked as image quality deviation areas, and image quality feedback data is generated.

7. The tracking method based on intelligent big data according to claim 6 is characterized in that: The shooting instructions of the step-by-step shooting instruction set are optimized according to the image quality feedback data to obtain an optimized shooting instruction set, including: The image quality feedback data is correlated and analyzed with the current step-by-step shooting instruction set, the instruction parameters causing the image quality deviation are extracted, and the parameter deviation data is generated; According to the parameter deviation data, the movement trajectory, shooting angle and focal length change of the corresponding device in the step-by-step shooting instruction set are adjusted to generate optimized parameter data; According to the optimization parameter data, the priority of the shooting instructions in each time period in the step-by-step shooting instruction set is reallocated to form an optimized shooting instruction set.

8. A tracking system based on intelligent big data, characterized in that: Applied to the method according to any one of claims 1 to 7, the system comprises: Data acquisition module, used to obtain real-time image data of the target object, scene environment parameter data and auxiliary behavior data; The image data processing module is used to perform frame processing on the real-time image data, extract the frames with significant changes in the picture, and generate key frame data; The environmental data processing module is used to classify and analyze the scene environmental parameter data and extract dynamic environmental characteristic data and static background characteristic data; A behavior data processing module is used to analyze the auxiliary behavior data, extract the movement pattern of the target object, and generate behavior characteristic data; A motion prediction module is used to calculate the preliminary position and motion trend of the target object based on the key frame data and the behavior characteristic data in combination with the dynamic environment characteristics, and generate motion prediction data; The area planning module is used to divide the activity range of the target object according to the motion prediction data and the static background characteristics, and generate the target area planning data; The shooting plan generation module is used to formulate shooting path data according to the target area planning data and generate a preliminary shooting plan, which includes the movement trajectory of the camera equipment, the shooting angle and the focal length change; A shooting plan refinement module is used to refine the preliminary shooting plan into a step-by-step shooting instruction set within a time period, and dynamically adjust the shooting instructions of the step-by-step shooting instruction set according to the real-time changes in the behavior of the target object; A shooting instruction execution module is used to execute the step-by-step shooting instruction set in real time and control the device to obtain continuous shooting images; The image quality feedback module is used to analyze the captured images in real time, evaluate the quality of the captured images, and generate image quality feedback data; A shooting instruction optimization module, used for optimizing the shooting instructions of the step-by-step shooting instruction set according to the image quality feedback data; The storage module is used to store the continuously shot images and related dynamic data as an expandable multi-dimensional data set.

9. A computing device, characterized in that include: one or more processors; A storage device for storing one or more programs, when the one or more programs are executed by the one or more processors, the one or more processors implement the method as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program, which, when executed by a processor, implements the method according to any one of claims 1 to 7.

Images