A large-scale mass sports event comprehensive evaluation system based on image recognition technology

Abstract

This invention relates to the field of data processing technology, specifically a comprehensive evaluation system for large-scale mass sports events based on image recognition technology. The system includes a time-frame synchronization module, a target tracking and detection module, a participation analysis module, a heatmap generation module, a trend prediction module, and a report generation module. First, it synchronizes multiple event video streams. Then, it tracks and detects the movement trajectories of participants, statistically analyzing the number of participants and their trajectory distribution to obtain real-time participation. The participation data is then processed into a grid to generate a dynamic heatmap, and the spatiotemporal evolution trend of participation concentration is predicted based on the event schedule. Finally, the real-time participation, dynamic heatmap, and prediction results are integrated to generate a comprehensive event evaluation report. This invention can improve the efficiency of generating reference information for the comprehensive evaluation of large-scale mass sports events based on image recognition technology.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a comprehensive evaluation system for large-scale mass sports events based on image recognition technology. Background Technology

[0002] In the comprehensive evaluation of large-scale mass sports events, existing technologies struggle to achieve precise and coordinated processing of multiple video streams from the event site. The lack of a unified time-base calibration mechanism during data acquisition leads to frame misalignment in video data from different sources, directly impacting the consistency and accuracy of participant target detection. Furthermore, existing target tracking technologies are prone to issues such as trajectory breaks and feature matching deviations when dealing with densely populated and complex event scenarios, failing to effectively acquire continuous motion trajectory data of participants. This results in significant errors in personnel counts and trajectory distribution feature extraction, making it difficult to generate valuable real-time participation data.

[0003] Existing event evaluation technologies have significant shortcomings in data integration and in-depth analysis. They cannot efficiently merge scattered participation-related data and transform it into an evaluation medium that reflects the dynamics of the event, making it difficult to intuitively present the spatial distribution differences and temporal patterns of participation. Furthermore, existing trend extrapolation methods fail to fully integrate the inherent correlation between the event schedule and real-time participation data, and their analysis of the spatiotemporal evolution of public participation concentration is insufficient, resulting in inaccurate predictions. This means that traditional comprehensive evaluation reports often only present a partial, static picture of the event, failing to comprehensively cover the core evaluation needs of each stage of the event and providing forward-looking and practical decision-making basis for event organization optimization and rational resource allocation. Summary of the Invention

[0004] To achieve the above objectives, the present invention provides a comprehensive evaluation system for large-scale mass sports events based on image recognition technology. The system comprises a time-frame synchronization module, a target tracking and detection module, a trajectory analysis module, a heatmap generation module, a trend analysis module, and a report generation module, wherein:

[0005] The time-frame synchronization module is used to perform time-frame synchronization processing on multiple video streams at the event site to obtain synchronized video data of the event site;

[0006] The target tracking and detection module is used to track and detect participants in synchronous video data to obtain continuous motion trajectory data of participants at the event site;

[0007] The participation analysis module is used to extract and statistically analyze the number of people and trajectory distribution characteristics at the event site based on continuous motion trajectory data, so as to obtain real-time participation data at the event site.

[0008] The heatmap generation module is used to perform spatial gridding processing on real-time participation data to obtain a dynamic heatmap of the event site;

[0009] The trend projection module is used to project the spatiotemporal evolution of the concentration of public participation at different stages of the event based on dynamic heat maps and the event schedule, and to obtain the prediction results for the event.

[0010] The report generation module is used to integrate and analyze real-time participation data, dynamic heatmaps, and prediction results to obtain a comprehensive evaluation report of the event.

[0011] In a preferred embodiment, when the time-frame synchronization module performs time-frame synchronization processing on multiple video streams from the event venue to obtain synchronized video data from the event venue, it is specifically used for:

[0012] The clocks of multiple video streams from the event site are synchronized to obtain a unified time reference for the event.

[0013] Based on a unified time reference, timestamps are applied to the data images of multiple video streams to obtain a sequence of video frames from the event.

[0014] Frame normalization is performed on the video frame sequence to obtain a set of video frames from the event.

[0015] The video frame set is time-series encapsulated to obtain synchronized video data from the event.

[0016] In a preferred embodiment, when the time-frame synchronization module performs frame normalization processing on the video frame sequence to obtain a set of video frames from the event, it is specifically used for:

[0017] The frames in the video frame sequence are timestamped to obtain the sequence of frames to be processed from the event scene;

[0018] The frame sequence to be processed is buffered to obtain the frame queue of the event scene;

[0019] Based on preset timing rules, the frame queue is reassembled in order to obtain a continuous video frame stream of the event.

[0020] Integrate continuous video frame streams to generate a set of video frames from the event.

[0021] In a preferred embodiment, when the target tracking and detection module performs participant tracking and detection on synchronized video data to obtain continuous motion trajectory data of participants at the event site, it is specifically used for:

[0022] Participant target detection is performed on synchronized video data to obtain a set of detection results for participants at the event site;

[0023] The apparent features and motion vectors in the detection result set are fused to obtain the participant feature descriptor;

[0024] Based on participant feature descriptors, multi-target data association and matching links are performed on the participant detection boxes to obtain trajectory fragments of the event site;

[0025] Motion trend smoothing is performed on the breaks in the trajectory segments to obtain continuous motion trajectory data of the event site.

[0026] In a preferred embodiment, when the participation analysis module extracts and statistically analyzes the number of people and trajectory distribution characteristics at the event site based on continuous motion trajectory data to obtain real-time participation data at the event site, it is specifically used for:

[0027] Based on continuous motion trajectory data, the number of unique participant identifiers in different zones of the event venue was deduplicated to obtain the statistical results of the number of people at the event venue.

[0028] Based on the position sequence in continuous motion trajectory data, path clustering analysis is performed on the participants' movement paths to obtain the trajectory distribution characteristics at the event site.

[0029] By integrating and analyzing the statistical results of the number of people with the characteristics of trajectory distribution, real-time participation data of the event site can be obtained.

[0030] In a preferred embodiment, when the participation analysis module performs path clustering analysis on the participants' movement paths based on the position sequence in continuous motion trajectory data to obtain the trajectory distribution characteristics of the event site, it is specifically used for:

[0031] Based on the location sequence, spatial density clustering is performed on the coordinate points of the movement paths of the participants to obtain the spatial clustering distribution of the movement paths.

[0032] By performing direction statistics on motion vectors in the position sequence, the mainstream pattern of motion vectors can be obtained;

[0033] The density distribution characteristics of movement paths are obtained by quantifying the number of people per unit area in spatially clustered regions. The calculation formula for this quantification is as follows:

[0034] ;

[0035] In the formula, Indicates the first Population density in a spatial cluster area Indicates the first The number of people in a spatial cluster area Indicates the first The area of ​​each spatial cluster region;

[0036] By integrating the spatial clustering area distribution, mainstream patterns, and density distribution characteristics, the trajectory distribution characteristics of the event site are obtained.

[0037] In a preferred embodiment, when the heatmap generation module performs spatial gridding processing on real-time participation data to obtain a dynamic heatmap of the event site, it is specifically used for:

[0038] The planar space of the event venue is divided into regular grids to obtain the spatial grid framework of the event venue;

[0039] The regional heat level and activity intensity index in the real-time participation data are gridded to obtain the gridded participation value distribution at the event site;

[0040] Based on the gridded participation value distribution, the values ​​of grid nodes in the event site are rendered by color mapping to obtain a participation heatmap of the event site.

[0041] By combining the participation heatmaps over time, a dynamic heatmap of the participation distribution at the event site is obtained.

[0042] In a preferred embodiment, when the trend extrapolation module performs trend extrapolation based on dynamic heatmaps and in conjunction with the event schedule to analyze the spatiotemporal evolution trend of public participation concentration at different stages of the event and obtain the predicted results for the event, it is specifically used for:

[0043] The event schedule was analyzed in a structured manner to obtain a phased time sequence framework for the event.

[0044] Based on dynamic heat maps, the spatial distribution and intensity changes of mass participation concentration at the event site are extracted in stages to obtain the staged spatiotemporal characteristics of mass participation concentration.

[0045] Based on the phased spatiotemporal characteristics, and combined with the trend fitting method, we conducted a correlation analysis on the characteristic changes of the concentration of mass participation, and obtained the spatiotemporal evolution law of the concentration of mass participation.

[0046] Based on the spatiotemporal evolution patterns, the subsequent distribution time points of mass participation concentration are extrapolated to obtain the predicted results for the event site.

[0047] In a preferred embodiment, when the trend extrapolation module performs correlation analysis on the characteristic changes of mass participation concentration based on phased spatiotemporal characteristics and combined with trend fitting methods to obtain the spatiotemporal evolution law of mass participation concentration, it is specifically used for:

[0048] Based on the phased spatiotemporal characteristics, the trend of change patterns in the concentration of mass participation is identified to obtain the temporal trend of mass participation concentration.

[0049] A correlation analysis was conducted on the spatial regional changes in the concentration of public participation to obtain the correlation characteristics between spatial regions;

[0050] Based on the temporal trends and correlation characteristics, the synergistic change patterns of mass participation concentration in the temporal and spatial dimensions are integrated and deduced to obtain the spatiotemporal evolution law of mass participation concentration.

[0051] In a preferred embodiment, when the report generation module performs a fusion analysis of real-time participation data, dynamic heatmaps, and prediction results to obtain a comprehensive evaluation report of the event, it is specifically used for:

[0052] By fusing real-time participation data, dynamic heatmaps, and prediction results, multi-dimensional evaluation data of the event site is obtained.

[0053] By comprehensively and quantitatively integrating the multi-dimensional evaluation data, the evaluation results of the event site are obtained;

[0054] Based on the quantitative evaluation results and the preset report template, the evaluation content of the quantitative evaluation results is analyzed and synthesized to obtain a comprehensive evaluation report of the event site.

[0055] Compared with the prior art, the present invention has the following beneficial effects:

[0056] 1. This invention constructs a unified time benchmark through time-frame synchronization processing to ensure the consistency of multi-channel video data. It obtains the continuous motion trajectory of participants by using feature fusion and trajectory smoothing technology, and accurately extracts the statistical results of the number of people and trajectory distribution characteristics by combining spatial density clustering and quantization calculation. This greatly improves the accuracy and comprehensiveness of real-time participation data and provides solid and rich data support for event evaluation.

[0057] 2. The system transforms real-time participation data into dynamic heatmaps, intuitively presenting the spatial distribution and temporal changes of participation. Through phased spatiotemporal feature extraction and trend fitting analysis, it accurately predicts the spatiotemporal evolution trend of mass participation concentration. Finally, it integrates multi-dimensional data to form a comprehensive evaluation report, effectively improving the intuitiveness and foresight of the event evaluation and providing scientific and practical decision support for the optimization and adjustment of each stage of the event. Attached Figure Description

[0058] Figure 1 A system architecture diagram of a comprehensive evaluation system for large-scale mass sports events based on image recognition technology, provided in an embodiment of the present invention;

[0059] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0060] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments belong to some, but not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0061] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “said” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0062] Depending on the context, the word "if" or "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0063] Furthermore, the timing of the steps in the following method embodiments is merely an example and not a strict limitation.

[0064] In practice, the server-side equipment deployed in a comprehensive evaluation system for large-scale mass sports events based on image recognition technology may consist of one or more devices. This system can be implemented as a business instance, a virtual machine, or hardware devices. For example, it can be implemented as a business instance deployed on one or more devices in a cloud node. Simply put, it can be understood as software deployed on a cloud node to provide a comprehensive evaluation system for large-scale mass sports events based on image recognition technology to various user terminals. Alternatively, it can be implemented as a virtual machine deployed on one or more devices in a cloud node, with application software installed to manage various user terminals. Or, it can also be implemented as a server composed of numerous identical or different types of hardware devices, with one or more hardware devices configured to provide a comprehensive evaluation system for large-scale mass sports events based on image recognition technology to various user terminals.

[0065] In terms of implementation, a comprehensive evaluation system for large-scale mass sports events based on image recognition technology and the user terminal are mutually compatible. Specifically, if the comprehensive evaluation system for large-scale mass sports events based on image recognition technology is implemented as an application installed on a cloud service platform, then the user terminal acts as a client establishing a communication connection with that application; or if the comprehensive evaluation system for large-scale mass sports events based on image recognition technology is implemented as a website, then the user terminal acts as a webpage; or if the comprehensive evaluation system for large-scale mass sports events based on image recognition technology is implemented as a cloud service platform, then the user terminal acts as a mini-program within an instant messaging application.

[0066] like Figure 1 The diagram shown is a system architecture diagram of a comprehensive evaluation system for large-scale mass sports events based on image recognition technology, provided by an embodiment of the present invention.

[0067] The comprehensive evaluation system 100 for large-scale mass sports events based on image recognition technology described in this invention can be set up in a cloud server. In terms of implementation, it can be used as one or more service devices, or as an application installed in the cloud (e.g., a mobile service operator's server, server cluster, etc.), or it can be developed into a website. Depending on the functions implemented, the comprehensive evaluation system 100 for large-scale mass sports events based on image recognition technology may include a time-frame synchronization module 101, a target tracking and detection module 102, a participation analysis module 103, a heatmap generation module 104, a trend inference module 105, and a report generation module 106. The modules described in this invention can also be called units, referring to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, stored in the memory of the electronic device.

[0068] In this embodiment of the invention, in a comprehensive evaluation system for large-scale mass sports events based on image recognition technology, each of the above-mentioned modules can be implemented independently and can call other modules. Here, "calling" can be understood as one module connecting to multiple modules of another type and providing corresponding services to those connected modules. The comprehensive evaluation system for large-scale mass sports events based on image recognition technology provided by this embodiment of the invention allows for adjustment of the applicability of the system architecture without modifying the program code. This is achieved by adding modules and directly calling them, enabling cluster-based horizontal expansion and flexibly expanding the system. In practical applications, the above modules can be set in the same device or different devices, or in virtual devices, such as service instances in a cloud server.

[0069] The following describes, with reference to specific embodiments, each component and its specific workflow of a comprehensive evaluation system for large-scale mass sports events based on image recognition technology:

[0070] The time-frame synchronization module 101 is used to perform time-frame synchronization processing on multiple video streams at the event site to obtain synchronized video data of the event site.

[0071] In this embodiment of the invention, when the time-frame synchronization module performs time-frame synchronization processing on multiple video streams from the event venue to obtain synchronized video data from the event venue, it is specifically used for:

[0072] The clocks of multiple video streams from the event site are synchronized to obtain a unified time reference for the event.

[0073] Based on a unified time reference, timestamps are applied to the data images of multiple video streams to obtain a sequence of video frames from the event.

[0074] Frame normalization is performed on the video frame sequence to obtain a set of video frames from the event.

[0075] The video frame set is time-series encapsulated to obtain synchronized video data from the event.

[0076] When the time-frame synchronization module performs frame normalization processing on the video frame sequence to obtain a set of video frames from the event, it is specifically used for:

[0077] The frames in the video frame sequence are timestamped to obtain the sequence of frames to be processed from the event scene;

[0078] The frame sequence to be processed is buffered to obtain the frame queue of the event scene;

[0079] Based on preset timing rules, the frame queue is reassembled in order to obtain a continuous video frame stream of the event.

[0080] Integrate continuous video frame streams to generate a set of video frames from the event.

[0081] Each acquisition device corresponding to a live video stream of the event is connected to the same GPS timing module. This module outputs a pulse signal every second. Each acquisition device uses this pulse signal as a reference to adjust the ticking rhythm of its internal clock, so that the clock time and ticking frequency of all acquisition devices are completely aligned. After calibration, the time provided by this GPS timing module is used as the unified time reference for processing multiple video streams at the event site.

[0082] Based on the established unified time reference, each video stream embeds the time information corresponding to the current unified time reference into the data header area of ​​the frame image while acquiring and generating images. During the embedding process, it is ensured that the time information and the image data correspond one-to-one without misalignment. After the time information of each frame image of all multiple video streams has been embedded, all image frames are integrated and arranged according to the time sequence of the unified time reference carried by each frame image to form a video frame sequence of the event site.

[0083] The video frame sequence is read frame by frame. The time reference time carried by two adjacent frames is compared. The acquisition time of a single frame is preset to 40 milliseconds. If the time interval between two adjacent frames is 40 milliseconds, the two frames are retained. If the time interval between two adjacent frames is 0 milliseconds, it means that there is a duplicate frame, and the duplicate image of the next frame is deleted. If the time interval between two adjacent frames is more than 40 milliseconds, it means that there is a lost frame, and a frame is added that is completely identical to the previous frame in content and whose time is the previous frame time plus 40 milliseconds. After all frame checking, deletion and addition operations are completed, the video frame set of the event site is obtained.

[0084] Based on the unified time reference time carried by each frame in the video frame set, the pixel data and embedded time information of each frame are extracted sequentially. The pixel data and time information of a single frame are bound together to form an independent frame data unit. All frame data units are concatenated in time order. At the same time, a format identifier is added to the header of the concatenated data stream to indicate that the frame interval of the data stream is 40 milliseconds and the time reference is GPS time. After completing the concatenation and adding the identifier, the synchronous video data of the event site is obtained.

[0085] Each frame in the video frame sequence is read frame by frame. The millisecond-level timestamps carried by each frame based on a unified time base are extracted. Using the timestamp of the first frame in the video frame sequence as the reference point, the timestamps of each subsequent frame are checked. If the interval between the timestamp of a frame and the timestamp of the previous frame is not equal to 40 milliseconds, the timestamp of that frame is adjusted to the value of the timestamp of the previous frame plus 40 milliseconds. After adjustment, the image data of that frame remains unchanged. After all frames have completed the timestamp check and adjustment, the sequence of frames to be processed at the event site is obtained.

[0086] A circular buffer with a fixed capacity of 10 frames is constructed. Frame images are stored into the buffer in the order of their timestamps according to the sequence of frames to be processed. When the buffer is full of 10 frames, the frame image with the earliest timestamp in the buffer is output first, and then the next frame image in the sequence of frames to be processed is stored into the buffer. The storage and output operations are performed in a loop to ensure that the output frame images are always continuous in the order of their timestamps, and finally form a frame queue of the event.

[0087] The preset timing rules are that the frames in the frame queue must be arranged in ascending order according to the timestamp of a unified time base, and the timetamp interval between adjacent frames is fixed at 40 milliseconds. The image frames in the frame queue are read frame by frame, and the timestamp difference between the current frame and the previous frame is checked. If the difference is 40 milliseconds, the order is kept unchanged. If the difference is 0 milliseconds, the current frame is removed. If the difference exceeds 40 milliseconds, a new frame with the same image content as the previous frame and a timestamp of the previous frame plus 40 milliseconds is added. After all checks and adjustments are completed, a continuous video frame stream of the event is obtained.

[0088] The image frames and corresponding timestamps in the continuous video frame stream are extracted frame by frame. All frames are traversed and checked. If there are frames with the same timestamp, only the first frame is kept and subsequent duplicate frames are removed. The remaining non-duplicate and sequentially connected frames are associated in order of their timestamps to form a complete time series. Each frame retains the original image data and the calibrated timestamp, and finally a video frame set of the event is generated.

[0089] The target tracking and detection module 102 is used to track and detect participants in synchronous video data to obtain continuous motion trajectory data of participants in the event.

[0090] In this embodiment of the invention, when the target tracking and detection module performs participant tracking and detection on synchronized video data to obtain continuous motion trajectory data of participants at the event site, it is specifically used for:

[0091] Participant target detection is performed on synchronized video data to obtain a set of detection results for participants at the event site;

[0092] The apparent features and motion vectors in the detection result set are fused to obtain the participant feature descriptor;

[0093] Based on participant feature descriptors, multi-target data association and matching links are performed on the participant detection boxes to obtain trajectory fragments of the event site;

[0094] Motion trend smoothing is performed on the breaks in the trajectory segments to obtain continuous motion trajectory data of the event site.

[0095] Synchronous video data is split into frames, and image information is extracted frame by frame. Participants and background in the image are distinguished by pixel-by-pixel comparison. The background area is referenced by fixed scene pixels of the event venue, and the participant area is the area with a difference from the background pixels and an area greater than 100 square pixels. A rectangular detection box is marked for each identified participant, and the selection range completely covers the participant area. The participant detection boxes and corresponding position information of all frames are integrated to obtain the detection result set of participants in the event venue.

[0096] The appearance features and motion vectors corresponding to the detection boxes of each participant are extracted from the detection results set. The appearance features include the color of the participant's clothing, the shape of the outline, and the relative position of key parts of the limbs. The motion vector is the position offset information between the detection box of the current frame and the corresponding detection box of the previous frame. The two types of features are associated in the order of appearance features first and motion vectors second. The feature information is uniformly encoded and stored. Each participant corresponds to a unique encoded feature information, which is the participant feature descriptor.

[0097] Extract the participant feature descriptors of all participants in the current frame and compare them one by one with the participant feature descriptors of all participants in the previous frame. During the comparison, prioritize checking the clothing color and outline shape in the appearance features. If the two are completely consistent, further confirm that the motion vector offset range is within 50 pixels allowed in a single frame image. If the above conditions are met, it is determined to be the same participant. Link the corresponding detection boxes. Detection boxes that do not match the corresponding descriptor are used as the starting point of the new trajectory. The trajectory that fails to match and has no new detection box to supplement is terminated. Integrate all the linked detection box sequences to obtain the trajectory segment of the event site.

[0098] Trajectory segments are traversed to identify trajectory breaks, i.e., locations where two consecutive frames lack corresponding detection boxes. The motion direction and position offset patterns of the trajectory segments in the three frames before the break are extracted. Virtual detection boxes are then added at the break points according to these patterns. The position of the virtual detection box is consistent with the offset distance and direction of the detection box in the last frame before the break, and with the offset between the two previous frames. After the addition is completed, the continuity of the trajectory segments is adjusted to ensure that the position transition of adjacent detection boxes conforms to the motion patterns, ultimately obtaining continuous motion trajectory data of the event site.

[0099] The participation analysis module 103 is used to extract and statistically analyze the number of people and trajectory distribution characteristics at the event site based on continuous motion trajectory data, so as to obtain real-time participation data at the event site.

[0100] In this embodiment of the invention, when the participation analysis module extracts and statistically analyzes the number of people and trajectory distribution characteristics at the event site based on continuous motion trajectory data to obtain real-time participation data at the event site, it is specifically used for:

[0101] Based on continuous motion trajectory data, the number of unique participant identifiers in different zones of the event venue was deduplicated to obtain the statistical results of the number of people at the event venue.

[0102] Based on the position sequence in continuous motion trajectory data, path clustering analysis is performed on the participants' movement paths to obtain the trajectory distribution characteristics at the event site.

[0103] By integrating and analyzing the statistical results of the number of people with the characteristics of trajectory distribution, real-time participation data of the event site can be obtained.

[0104] When the participation analysis module performs path clustering analysis on the participants' movement paths based on the position sequence in continuous motion trajectory data to obtain the trajectory distribution characteristics of the event site, it is specifically used for:

[0105] Based on the location sequence, spatial density clustering is performed on the coordinate points of the movement paths of the participants to obtain the spatial clustering distribution of the movement paths.

[0106] By performing direction statistics on motion vectors in the position sequence, the mainstream pattern of motion vectors can be obtained;

[0107] The density distribution characteristics of movement paths are obtained by quantifying the number of people per unit area in spatially clustered regions. The calculation formula for this quantification is as follows:

[0108] ;

[0109] In the formula, Indicates the first Population density in a spatial cluster area Indicates the first The number of people in a spatial cluster area Indicates the first The area of ​​each spatial cluster region;

[0110] By integrating the spatial clustering area distribution, mainstream patterns, and density distribution characteristics, the trajectory distribution characteristics of the event site are obtained.

[0111] First, the boundaries of the event venue are defined, with each zone corresponding to a fixed spatial range. The boundaries are based on the physical markers of the actual event venue plan, ensuring that the ranges of each zone do not overlap or have any omissions. Then, the participant identifiers in each zone at each time point are extracted from the continuous motion trajectory data. Each participant identifier is a unique identification code for each independent participant. Subsequently, all the participant identifiers extracted in each zone are compared one by one, and duplicate identifiers are removed, retaining only the unique identifiers that appear only once. Finally, the total number of unique identifiers in each zone is counted. The total number of unique identifiers in all zones and the corresponding number in each zone together constitute the statistical result of the number of people at the event venue.

[0112] The position sequence of each independent participant is extracted from continuous motion trajectory data. The position sequence records the specific location of the participant at each minute in chronological order. The location is based on the preset spatial coordinate points of the event site. Each coordinate point corresponds to a unique physical location in the venue. Then, all participants' position sequences are classified according to the position change pattern corresponding to the time sequence. That is, if the position coordinate points of two participants' position sequences are completely consistent at each minute, they are classified into the same category. Each category forms a typical movement path. At the same time, the number of participants corresponding to each type of path and the range of event site coordinate points covered by each type of path are counted. These typical movement paths, the number of participants corresponding to each type of path, and the range of event site coordinate points covered by each type of path together constitute the trajectory distribution characteristics of the event site.

[0113] First, we analyze the total number of unique identifiers for each zone in the personnel statistics, as well as the typical movement paths, corresponding number of participants, and coverage coordinate range in the trajectory distribution characteristics. Then, we correlate the total number of unique identifiers for each zone with the typical movement paths covered within that zone to determine the proportion of participants for each typical movement path within the corresponding zone to the total number of unique identifiers for that zone. At the same time, we confirm whether the coordinate range covered by each typical movement path is completely within the corresponding zone. The results of both are combined to form the participation status data for each zone. After integrating the participation status data of all zones, we form real-time participation data of the event site that can reflect the degree of matching between the personnel gathering situation and movement patterns in each area.

[0114] Complete position sequences are collected from the continuous movement trajectory data of all participants. All coordinate points contained in each position sequence are extracted and projected onto a unified spatial coordinate system of the event site. The coordinate system is divided into the smallest spatial grid with 1 meter units. The grid range to which each coordinate point belongs is determined one by one. The number of coordinate points contained in each grid is counted. Grids with 5 or more adjacent coordinate points are merged into a whole area. At the same time, the boundary range of each merged area is determined. The boundary range is based on the polygon formed by the outermost coordinate points in the area. All the merged areas formed in the end are the spatial clustering areas of the movement path.

[0115] Extract the motion vector formed by two adjacent coordinate points in the position sequence of each participant. With the north of the coordinate system as the 0-degree reference, calculate the direction angle corresponding to each motion vector clockwise. Divide the direction angle into 24 fixed intervals of 15 degrees each. Record the angle interval corresponding to each motion vector one by one, count the number of motion vectors contained in each interval, compare the number of motion vectors in the 24 intervals, and determine the direction range corresponding to the interval with the most motion vectors as the mainstream pattern of motion vectors. The mainstream pattern should be clearly marked with the corresponding angle range.

[0116] For each identified spatial cluster area, the number of participants within that area is counted one by one. The counting is based on whether the participant's coordinates fall within the area boundary. Only participants whose coordinates are completely within the area are counted to obtain the number of people in each spatial cluster area. The area of ​​each spatial cluster area is determined by measuring the area of ​​the boundary polygon. During the measurement, the number of complete grids contained within the polygon is calculated in units of coordinate system grid. Any fractions less than one complete grid are proportionally reduced and then accumulated to obtain the actual area of ​​each area. The number of people in each spatial cluster area is divided by the area of ​​that area to obtain the population density of each spatial cluster area. The population density of all areas and the corresponding area information are summarized to form the density distribution characteristics of the movement path.

[0117] The spatial clustering distribution includes the boundary and location information of each area, the angle interval information corresponding to the mainstream motion vector pattern, and the population density information of each area included in the density distribution characteristics. All of these are entered into the same analysis table. The table is filled in one by one according to the spatial clustering area to ensure that the clustering range, mainstream direction, and population density data of each area are accurately matched without any misalignment or omission. The table is integrated to realize the interrelation of the three types of features. Finally, the participant clustering situation, main movement direction, and population density in different areas of the event site are extracted from the integrated information to form the trajectory distribution characteristics of the event site.

[0118] The heat map generation module 104 is used to perform spatial gridding processing on real-time participation data to obtain a dynamic heat map of the event site.

[0119] In this embodiment of the invention, when the heatmap generation module performs spatial gridding processing on real-time participation data to obtain a dynamic heatmap of the event site, it is specifically used for:

[0120] The planar space of the event venue is divided into regular grids to obtain the spatial grid framework of the event venue;

[0121] The regional heat level and activity intensity index in the real-time participation data are gridded to obtain the gridded participation value distribution at the event site;

[0122] Based on the gridded participation value distribution, the values ​​of grid nodes in the event site are rendered by color mapping to obtain a participation heatmap of the event site.

[0123] By combining the participation heatmaps over time, a dynamic heatmap of the participation distribution at the event site is obtained.

[0124] Obtain the geographic boundary coordinates of the event site's planar space. These coordinates are based on the coordinates of the outer contour vertices of the actual event site. Determine the maximum horizontal and vertical coverage of the planar space. According to the preset grid side length standard, the horizontal coverage of the planar space is evenly divided according to the side length standard, and then the vertical coverage is evenly divided according to the same side length standard. The horizontal and vertical dividing lines intersect each other to form multiple square grids of the same size that do not overlap. Each grid is assigned a unique spatial identifier, which includes the grid's horizontal and vertical arrangement number. After all the square grids are arranged according to their actual positions, the spatial grid framework of the event site is obtained.

[0125] The criteria for classifying regional heat levels in real-time participation data were determined. This standard is based on the saturation of the number of people within a region, divided into five levels, from level one to level five, corresponding to saturation from 0% to 100%. Simultaneously, the criteria for determining activity intensity indicators were determined. This standard is based on the update frequency of personnel movement paths within a region, divided into five levels, from level one to level five, corresponding to update frequencies from 0 times per second to 5 times per second. Each statistical region in the real-time participation data was matched with a grid in the spatial grid framework. The matching basis was the overlap of the geographical coverage of the region and the grid. A grid with a 100% overlap corresponded to a unique statistical region. The heat level and activity intensity indicators of the corresponding statistical region were integrated. The integration method was to take the average of the two indicators as the participation value of that grid. The participation values ​​of all grids were summarized according to spatial identifiers to obtain the gridded participation value distribution of the event site.

[0126] A pre-defined correspondence between participation values ​​and colors is established, which is set as a color gradient sequence. The participation value, from lowest to highest, corresponds to a color gradient from blue to red, where the lowest value corresponds to pure blue, the middle value corresponds to yellow, and the highest value corresponds to pure red. Each grid in the gridded participation value distribution is traversed, and the participation value of each grid is extracted. The value is then matched with the pre-defined color gradient sequence to determine the color corresponding to each grid. According to the actual position of the grid in the spatial grid framework, each grid is filled with the matched color. After filling, the grids are integrated according to their spatial arrangement order to obtain the participation heatmap of the event site.

[0127] A time interval standard for time-series combination is determined. This standard is set as a fixed duration based on the time step of data collection at the event site. Participation heatmaps at different times are extracted according to this time interval. The number of extracted heatmaps is no less than ten to ensure that the change process of participation can be fully presented. All extracted participation heatmaps are arranged in chronological order of collection time, and the arrangement order is completely consistent with the time flow order. The playback rate of the heatmaps is set to two heatmaps per second. The arranged heatmaps are continuously displayed and combined at the set rate. During the combination process, the time mark of each heatmap is retained. Finally, a dynamic heatmap of participation distribution at the event site is obtained.

[0128] The trend projection module 105 is used to project the spatiotemporal evolution trend of mass participation concentration at different stages of the event based on dynamic heat maps and the event schedule, and to obtain the prediction results of the event.

[0129] In this embodiment of the invention, when the trend extrapolation module performs trend extrapolation based on dynamic heatmaps and in conjunction with the event schedule to analyze the spatiotemporal evolution trend of public participation concentration at different stages of the event and obtain the prediction results for the event site, it is specifically used for:

[0130] The event schedule was analyzed in a structured manner to obtain a phased time sequence framework for the event.

[0131] Based on dynamic heat maps, the spatial distribution and intensity changes of mass participation concentration at the event site are extracted in stages to obtain the staged spatiotemporal characteristics of mass participation concentration.

[0132] Based on the phased spatiotemporal characteristics, and combined with the trend fitting method, we conducted a correlation analysis on the characteristic changes of the concentration of mass participation, and obtained the spatiotemporal evolution law of the concentration of mass participation.

[0133] Based on the spatiotemporal evolution patterns, the subsequent distribution time points of mass participation concentration are extrapolated to obtain the predicted results for the event site.

[0134] The trend projection module, when performing correlation analysis on the characteristic changes of mass participation concentration based on phased spatiotemporal characteristics and combined with trend fitting methods to obtain the spatiotemporal evolution law of mass participation concentration, is specifically used for:

[0135] Based on the phased spatiotemporal characteristics, the trend of change patterns in the concentration of mass participation is identified to obtain the temporal trend of mass participation concentration.

[0136] A correlation analysis was conducted on the spatial regional changes in the concentration of public participation to obtain the correlation characteristics between spatial regions;

[0137] Based on the temporal trends and correlation characteristics, the synergistic change patterns of mass participation concentration in the temporal and spatial dimensions are integrated and deduced to obtain the spatiotemporal evolution law of mass participation concentration.

[0138] Collect the complete schedule documents released by the official event organizers, extract the specific start and end times, venues, event types, and session numbers for all events, and organize all events in chronological order. Divide the entire event cycle into several continuous and non-overlapping phases, with each phase containing only a single event or a group of related events held consecutively at the same venue. Clarify the start and end points and core event content of each phase, and organize the time information, corresponding events, and venues of each phase into a structured table. Each row in the table corresponds to a phase, and the columns include the phase time range, core events, and venue, ultimately forming a phased time sequence framework.

[0139] By deploying fixed cameras at key locations at the event site and collecting footage every 10 minutes to generate dynamic heat maps, the heat maps use the actual geographical layout of the site as a base map. Each camera's footage corresponds to a fixed area on the base map. Each heat map is broken down into several statistical units of equal area matching the base map. The number of people in each statistical unit is counted, and the number of people in each statistical unit is used as the basis for converting it into a heat value. The more people there are, the higher the heat value. Combined with a phased time series framework, heat map data for all time points within each phase is extracted, and the heat value of each statistical unit at each time point within that phase is recorded. At the same time, the specific geographical location of each statistical unit is marked, and the data is integrated to form a dataset containing heat values ​​for each phase, region, and time point, thus obtaining the phased spatiotemporal characteristics of the concentration of public participation.

[0140] This study analyzes the heat value data in the phased spatiotemporal characteristics of mass participation concentration. For each statistical unit, the heat value changes across different phases are analyzed. The heat values ​​of the same statistical unit in each phase are arranged chronologically. The heat values ​​of the same statistical unit in two adjacent phases are compared, and the difference between the heat values ​​of the later and earlier phases is calculated. Simultaneously, the heat values ​​of different statistical units within the same phase are compared to determine the ranking of heat values ​​for each statistical unit in that phase. Combining the event content and venue conditions of the corresponding phase, the study analyzes the correspondence between the heat value differences and rankings and these factors. This leads to the summary of the heat value change path of the same area in different phases, as well as the heat value distribution characteristics of different areas within the same phase, thus revealing the spatiotemporal evolution pattern of mass participation concentration.

[0141] Based on the spatiotemporal evolution of public participation concentration, the distribution time points to be extrapolated are determined. Each subsequent time point falls within the subsequent stage of the phased time series framework. For each subsequent time point, the core event content and venue conditions of the corresponding stage are referenced. The change path of public participation concentration in the corresponding stage under the same or similar event content and venue conditions in history is compared. Combined with the change of heat value of the same statistical unit in the currently completed stage, the heat value of each statistical unit at the subsequent time point is judged. At the same time, the difference in heat value change in the same area in adjacent stages is referenced to adjust the judgment results of the heat value of each statistical unit to ensure that the judgment results conform to the spatiotemporal evolution. The heat values ​​and corresponding geographical locations of each statistical unit at all subsequent time points are compiled and summarized to clarify the number and distribution of the public in each area at each subsequent time point, and the prediction results of the event site are obtained.

[0142] First, the overall research period is divided into several consecutive stages at fixed time intervals. Raw data on the concentration of public participation within the corresponding spatiotemporal range are extracted for each stage. The data must cover the participation status at each time node and corresponding spatial location. Then, for the concentration data within each stage, the numerical changes are analyzed node by node. At the same time, the differences in concentration values ​​at the same time node in adjacent stages are compared to identify the trend of concentration value changes within each stage and the connection between the trends of concentration changes between different stages. This completes the trend identification of the change pattern in the concentration of public participation and obtains the temporal trend of the concentration of public participation.

[0143] The research scope was divided into several independent spatial regions according to administrative boundaries. The concentration of public participation in each region was extracted at each time point during the research period. Then, taking a single region as a benchmark, the concentration value changes of that region and other regions at the same time point were compared one by one. The concentration change status of the corresponding region when the concentration of the benchmark region increased and the concentration change status of the corresponding region when the concentration of the benchmark region decreased were recorded. At the same time, the frequency of overlap of the above-mentioned change status between different regions was counted. The combination of regions with an overlap frequency of half or more of the total number of time points in the research period was determined to have a correlation, thereby clarifying the type of correlation between regions and obtaining the correlation characteristics between spatial regions.

[0144] Using the temporal trend of mass participation concentration as a framework, this study matches spatial region combinations with related characteristics within each time period, analyzes the coordinated changes in concentration of the same regional combination at different time stages, specifically whether the concentration of each region within the related regional combination follows the same trend when the temporal trend is upward, and whether the concentration of each region within the related regional combination follows the same trend when the temporal trend is downward. Furthermore, it integrates the coordinated changes in concentration of different related regional combinations within the same time period, forming a complete logic covering the changes in the temporal dimension and spatial regional relationships, thus revealing the spatiotemporal evolution law of mass participation concentration.

[0145] The report generation module 106 is used to integrate and analyze real-time participation data, dynamic heat maps, and prediction results to obtain a comprehensive evaluation report of the event.

[0146] In this embodiment of the invention, when the report generation module performs a fusion analysis of real-time participation data, dynamic heatmaps, and prediction results to obtain a comprehensive evaluation report of the event, it is specifically used for:

[0147] By fusing real-time participation data, dynamic heatmaps, and prediction results, multi-dimensional evaluation data of the event site is obtained.

[0148] By comprehensively and quantitatively integrating the multi-dimensional evaluation data, the evaluation results of the event site are obtained;

[0149] Based on the quantitative evaluation results and the preset report template, the evaluation content of the quantitative evaluation results is analyzed and synthesized to obtain a comprehensive evaluation report of the event site.

[0150] Using the spatial grid markers and timeline of the event site as a unified reference, the spatial grid adopts the previously defined square grid and unique identifier. The timeline covers the historical time stages corresponding to the real-time participation data, continuous time slices of the dynamic heatmap, and subsequent time nodes corresponding to the prediction results, ensuring complete connection of the three types of data in the spatiotemporal dimensions. The heat level and activity intensity values ​​of each grid in the real-time participation data are extracted and integrated. The color mapping of each grid in the dynamic heatmap is reversed to convert it into the corresponding mass participation concentration value. Simultaneously, the concentration prediction values ​​and spatial distribution characteristics of each grid at each subsequent time node in the prediction results are extracted. Based on the correspondence of "same grid - same time node," the three types of data are correlated and matched. Under the same grid and the same historical time node, the real-time participation value is correlated with the heatmap concentration value. Under the same grid and subsequent time nodes, the correlation between the predicted concentration value and the changes in historical real-time data is correlated. Grid time combinations with no corresponding data for three consecutive time nodes are eliminated. The average value of adjacent grids at the same time node is used to supplement the missing data, ultimately forming a full-dimensional data set covering historical real-time data, current heatmap characteristics, and subsequent prediction information—that is, multi-dimensional evaluation data of the event site.

[0151] Three core evaluation dimensions are established: spatial distribution adaptability, temporal stability, and subsequent trend controllability. Each dimension corresponds to clear judgment criteria. Spatial distribution adaptability focuses on the matching of grid participation values ​​with their functional area positioning; temporal stability focuses on the numerical fluctuations of the same grid at different time points; and subsequent trend controllability focuses on the fit between the predicted results and the site's carrying capacity. For spatial distribution adaptability, the number of grids with participation values ​​reaching level four or above in each of the competition area, spectator area, and service area is counted. The proportion of high-value grids in each functional area is judged to meet preset standards: no more than 30% in the competition area, no more than 60% in the spectator area, and no more than 20% in the service area. For temporal stability, the difference in participation values ​​between adjacent time points within the same grid is calculated. A difference controlled within level one of the heat levels indicates stability. The proportion of stable grids in the entire grid is then calculated. For subsequent trend controllability, the total number of high-value grids at each time point in the predicted results is checked to see if it exceeds the maximum carrying capacity of the site. This number is calculated by summing 70% of the total number of grids in each functional area. The results of the assessments across the various dimensions are integrated according to the following proportions: spatial distribution adaptability (40%), temporal stability (30%), and subsequent trend controllability (30%). The assessment results of each dimension are then summarized to clarify the achievement status and core issues of each dimension and form the on-site evaluation results of the event.

[0152] The pre-set report template includes five fixed chapters: Data Overview, Spatial Distribution Assessment, Temporal Dynamic Analysis, Subsequent Trend Judgment, and Comprehensive Conclusions and Recommendations. Each chapter corresponds to a clearly defined scope of content. The Data Overview chapter explains the coverage, spatiotemporal dimension division, and data completeness of the multi-dimensional assessment data. The Spatial Distribution Assessment chapter corresponds to the assessment results of spatial distribution adaptability. The Temporal Dynamic Analysis chapter corresponds to the assessment results of the stability of temporal changes. The Subsequent Trend Judgment chapter corresponds to the assessment results of the controllability of subsequent trends. The Comprehensive Conclusions and Recommendations chapter integrates the overall assessment and proposes targeted measures. Specific data, compliance status, and core issues from the quantitative assessment results are filled into the corresponding chapters one by one. Analytical explanations are provided simultaneously during the filling process. For example, the Spatial Distribution Assessment chapter explains the reasons for whether the proportion of high-value grids meets or exceeds the standard, and explains whether it meets the actual needs in the context of the event. The Subsequent Trend Judgment chapter identifies the specific time points and grid areas exceeding the carrying capacity and analyzes the potential impact. It is ensured that the content of each chapter is logically consistent with the preceding data and assessment results, the language conforms to technical report specifications, and there is no redundant information. Finally, the content of each chapter is integrated to form a logically coherent and information-complete comprehensive assessment report of the event.

[0153] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0154] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0155] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A comprehensive evaluation system for large-scale mass sports events based on image recognition technology, characterized in that, The system includes a time-frame synchronization module, a target tracking and detection module, an participation analysis module, a heatmap generation module, a trend inference module, and a report generation module, wherein: The time-frame synchronization module is used to perform time-frame synchronization processing on multiple video streams at the event site to obtain synchronized video data of the event site; The target tracking and detection module is used to track and detect participants in synchronous video data to obtain continuous motion trajectory data of participants at the event site; The participation analysis module is used to extract and statistically analyze the number of people and trajectory distribution characteristics at the event site based on continuous motion trajectory data, so as to obtain real-time participation data at the event site. The heatmap generation module is used to perform spatial gridding processing on real-time participation data to obtain a dynamic heatmap of the event site; The trend projection module is used to project the spatiotemporal evolution of the concentration of public participation at different stages of the event based on dynamic heat maps and the event schedule, and to obtain the prediction results for the event. The report generation module is used to integrate and analyze real-time participation data, dynamic heatmaps, and prediction results to obtain a comprehensive evaluation report of the event.

2. The comprehensive evaluation system for large-scale mass sports events based on image recognition technology as described in claim 1, characterized in that, When the time-frame synchronization module performs time-frame synchronization processing on multiple video streams from the event site to obtain synchronized video data from the event site, it is specifically used for: The clocks of multiple video streams from the event site are synchronized to obtain a unified time reference for the event. Based on a unified time reference, timestamps are applied to the data images of multiple video streams to obtain a sequence of video frames from the event. Frame normalization is performed on the video frame sequence to obtain a set of video frames from the event. The video frame set is time-series encapsulated to obtain synchronized video data from the event.

3. The comprehensive evaluation system for large-scale mass sports events based on image recognition technology as described in claim 2, characterized in that, When the time-frame synchronization module performs frame normalization processing on the video frame sequence to obtain a set of video frames from the event, it is specifically used for: The frames in the video frame sequence are timestamped to obtain the sequence of frames to be processed from the event scene; The frame sequence to be processed is buffered to obtain the frame queue of the event scene; Based on preset timing rules, the frame queue is reassembled in order to obtain a continuous video frame stream of the event. Integrate continuous video frame streams to generate a set of video frames from the event.

4. The comprehensive evaluation system for large-scale mass sports events based on image recognition technology as described in claim 1, characterized in that, When the target tracking and detection module performs participant tracking and detection on synchronized video data to obtain continuous motion trajectory data of participants at the event site, it is specifically used for: Participant target detection is performed on synchronized video data to obtain a set of detection results for participants at the event site; The apparent features and motion vectors in the detection result set are fused to obtain the participant feature descriptor; Based on participant feature descriptors, multi-target data association and matching links are performed on the participant detection boxes to obtain trajectory fragments of the event site; Motion trend smoothing is performed on the breaks in the trajectory segments to obtain continuous motion trajectory data of the event site.

5. The comprehensive evaluation system for large-scale mass sports events based on image recognition technology as described in claim 1, characterized in that, When the participation analysis module extracts and statistically analyzes the number of people and trajectory distribution characteristics at the event site based on continuous motion trajectory data to obtain real-time participation data, it is specifically used for: Based on continuous motion trajectory data, the number of unique participant identifiers in different zones of the event venue was deduplicated to obtain the statistical results of the number of people at the event venue. Based on the position sequence in continuous motion trajectory data, path clustering analysis is performed on the participants' movement paths to obtain the trajectory distribution characteristics at the event site. By integrating and analyzing the statistical results of the number of people with the characteristics of trajectory distribution, real-time participation data of the event site can be obtained.

6. The comprehensive evaluation system for large-scale mass sports events based on image recognition technology as described in claim 5, characterized in that, When the participation analysis module performs path clustering analysis on the participants' movement paths based on the position sequence in continuous motion trajectory data to obtain the trajectory distribution characteristics of the event site, it is specifically used for: Based on the location sequence, spatial density clustering is performed on the coordinate points of the movement paths of the participants to obtain the spatial clustering distribution of the movement paths. By performing direction statistics on motion vectors in the position sequence, the mainstream pattern of motion vectors can be obtained; The density distribution characteristics of movement paths are obtained by quantifying the number of people per unit area in spatially clustered regions. The calculation formula for this quantification is as follows: ； In the formula, Indicates the first Population density in a spatial cluster area Indicates the first The number of people in a spatial cluster area Indicates the first The area of ​​each spatial cluster region; By integrating the spatial clustering area distribution, mainstream patterns, and density distribution characteristics, the trajectory distribution characteristics of the event site are obtained.

7. The comprehensive evaluation system for large-scale mass sports events based on image recognition technology as described in claim 1, characterized in that, When the heatmap generation module performs spatial gridding processing on real-time participation data to obtain a dynamic heatmap of the event site, it is specifically used for: The planar space of the event venue is divided into regular grids to obtain the spatial grid framework of the event venue; The regional heat level and activity intensity index in the real-time participation data are gridded to obtain the gridded participation value distribution at the event site; Based on the gridded participation value distribution, the values ​​of grid nodes in the event site are rendered by color mapping to obtain a participation heatmap of the event site. By combining the participation heatmaps over time, a dynamic heatmap of the participation distribution at the event site is obtained.

8. A comprehensive evaluation system for large-scale mass sports events based on image recognition technology as described in claim 1, characterized in that, The trend projection module, when performing trend projection based on dynamic heatmaps and the event schedule to analyze the spatiotemporal evolution of public participation concentration at different stages of the event and obtain prediction results for the event, is specifically used for: The event schedule was analyzed in a structured manner to obtain a phased time sequence framework for the event. Based on dynamic heat maps, the spatial distribution and intensity changes of mass participation concentration at the event site are extracted in stages to obtain the staged spatiotemporal characteristics of mass participation concentration. Based on the phased spatiotemporal characteristics, and combined with the trend fitting method, we conducted a correlation analysis on the characteristic changes of the concentration of mass participation, and obtained the spatiotemporal evolution law of the concentration of mass participation. Based on the spatiotemporal evolution patterns, the subsequent distribution time points of mass participation concentration are extrapolated to obtain the predicted results for the event site.

9. A comprehensive evaluation system for large-scale mass sports events based on image recognition technology as described in claim 8, characterized in that, The trend projection module, when performing correlation analysis on the characteristic changes of mass participation concentration based on phased spatiotemporal characteristics and combined with trend fitting methods to obtain the spatiotemporal evolution law of mass participation concentration, is specifically used for: Based on the phased spatiotemporal characteristics, the trend of change patterns in the concentration of mass participation is identified to obtain the temporal trend of mass participation concentration. A correlation analysis was conducted on the spatial regional changes in the concentration of public participation to obtain the correlation characteristics between spatial regions; Based on the temporal trends and correlation characteristics, the synergistic change patterns of mass participation concentration in the temporal and spatial dimensions are integrated and deduced to obtain the spatiotemporal evolution law of mass participation concentration.

10. A comprehensive evaluation system for large-scale mass sports events based on image recognition technology as described in claim 1, characterized in that, When the report generation module performs a fusion analysis of real-time participation data, dynamic heatmaps, and prediction results to obtain a comprehensive evaluation report of the event, it is specifically used for: By fusing real-time participation data, dynamic heatmaps, and prediction results, multi-dimensional evaluation data of the event site is obtained. By comprehensively and quantitatively integrating the multi-dimensional evaluation data, the evaluation results of the event site are obtained; Based on the quantitative evaluation results and the preset report template, the evaluation content of the quantitative evaluation results is analyzed and synthesized to obtain a comprehensive evaluation report of the event site.

Images