Dynamic bright spot target statistics and analysis method based on visual tracking

Abstract

The invention relates to the technical field of visual target analysis, and discloses a dynamic bright spot target statistics and analysis method based on visual tracking. According to the method, a reference video sequence is collected to construct a background model database, a real-time video stream data set is obtained, a visual bright spot target is detected, and an average brightness value is calculated as a detection reference threshold value. And a visual bright spot target is analyzed in combination with the background model database, and a potential abnormal bright spot target is identified and marked. And when a potential abnormal bright spot target exists, extracting spatial position information and movement track characteristics of the potential abnormal bright spot target, and judging whether the potential abnormal bright spot target is a real abnormal bright spot target. And after real abnormity is determined, calculating an initial risk assessment value according to the occurrence frequency, extracting morphological contour and size change features, and performing similarity grouping analysis by using a feature matching algorithm. And judging whether a dynamic flicker bright spot target exists or not according to an analysis result, if so, correcting an initial risk assessment value based on a flicker mode characteristic setting adjustment factor, and adjusting subsequent dynamic tracking frame rate setting according to the corrected value.

Description

Technical Field

[0001] The present invention relates to the technical field of visual target analysis, and in particular to a dynamic bright spot target statistics and analysis method based on visual tracking. Background Art

[0002] Capturing and analyzing dynamic bright spot targets is a common task in various scenarios requiring visual monitoring. Existing technologies often use preset thresholds for bright spot target detection. These thresholds work well when ambient light levels are stable, but they lose their adaptability when light levels fluctuate, particularly during daytime and during cloud cover. A threshold that is too high above the actual light level will filter out bright spots that should have been detected, while a threshold that is too low will misidentify stray light in the background as bright spots, resulting in a large amount of invalid information.

[0003] The frame rate selection during tracking is also limited. Most solutions operate at a fixed frame rate. When the number of bright spots within the monitoring range increases suddenly or the speed of motion accelerates, a fixed frame rate cannot capture enough target details within a unit of time, resulting in broken motion trajectories. In scenes with sparse bright spots and slow motion, a fixed frame rate will continue to consume excessive computing resources, resulting in unnecessary consumption.

[0004] Existing methods for identifying abnormal bright spots often rely on a single feature, such as determining whether the brightness exceeds the normal range. However, in real-world scenarios, some bright spots are bright but within normal motion paths, while others exhibit irregular motion despite not reaching the threshold brightness. Relying solely on a single feature can lead to biased abnormality judgments. Furthermore, some bright spots flicker periodically, and the impact of this dynamic characteristic on the degree of abnormality is not taken into account, resulting in a discrepancy between the final analysis results and the actual situation. Summary of the Invention

[0005] The purpose of the present invention is to provide a dynamic bright spot target statistics and analysis method based on visual tracking to solve the problems raised in the above background technology.

[0006] To achieve the above-mentioned object, the present invention provides a method for counting and analyzing dynamic bright spot targets based on visual tracking, the method comprising:

[0007] Acquire a reference video sequence and construct a background model database; obtain a real-time video stream data set, determine a dynamic tracking frame rate based on the data size of the real-time video stream data set, detect all visual bright spot targets in the real-time video stream and calculate the average brightness value of the visual bright spot targets; set the average brightness value as a detection reference threshold; analyze and process the visual bright spot targets based on the detection reference threshold and the background model database, and identify and mark potential abnormal bright spot targets;

[0008] When there are potential abnormal bright spot targets, extract the spatial position information and motion trajectory characteristics of each potential abnormal bright spot target; and determine whether the potential abnormal bright spot target is a real abnormal bright spot target based on the spatial position information and motion trajectory characteristics;

[0009] When a target is determined to be a real abnormal bright spot, the initial risk assessment value is calculated based on the frequency of the abnormal bright spot target; the morphological contour features and size change features of the real abnormal bright spot target are extracted; and a feature matching algorithm is used to perform similarity grouping analysis on all real abnormal bright spot targets;

[0010] Determine whether there is a dynamic flickering bright spot target based on the results of the similarity grouping analysis; when it is determined that there is a dynamic flickering bright spot target, determine an adjustment factor based on the flickering pattern characteristics of the dynamic flickering bright spot target to correct the initial risk assessment value; and adjust the subsequent dynamic tracking frame rate setting based on the corrected risk assessment value.

[0011] Preferably, when determining the dynamic tracking frame rate based on the data volume of the real-time video stream data set, the following operations are performed: the data volume is compared with a preset low data volume threshold and a preset high data volume threshold respectively; the corresponding dynamic tracking frame rate configuration is output according to the comparison result; wherein the preset low data volume threshold is lower than the preset high data volume threshold; when the data volume does not exceed the preset low data volume threshold, the dynamic tracking frame rate is set to a low frame rate mode; when the data volume exceeds the preset low data volume threshold but does not exceed the preset high data volume threshold, the dynamic tracking frame rate is set to a medium frame rate mode; when the data volume exceeds the preset high data volume threshold, the dynamic tracking frame rate is set to a high frame rate mode; the frame rate value corresponding to the low frame rate mode is lower than that of the medium frame rate mode, and the frame rate value corresponding to the medium frame rate mode is lower than that of the high frame rate mode.

[0012] Preferably, when analyzing and processing the visual bright spot target based on the detection benchmark threshold and the background model database, the following operations are performed: the brightness values ​​of all visual bright spot targets in the real-time video stream are compared with the detection benchmark threshold; and the morphological features of each visual bright spot target are matched with the standard morphological features in the background model database; potential abnormal bright spot targets are identified and marked based on the results of the comparison and matching checks; when the brightness value of the visual bright spot target exceeds the ratio range set by the detection benchmark threshold, the visual bright spot target is determined to be a potential abnormal bright spot target and marked; when the morphological features of the visual bright spot target do not find a matching record in the background model database, the visual bright spot target is determined to be a potential abnormal bright spot target and marked.

[0013] Preferably, when judging whether a potential abnormal bright spot target is a real abnormal bright spot target based on the spatial position information and motion trajectory characteristics, the following operations are performed: when the spatial position information of the potential abnormal bright spot target shows that its motion trajectory deviates from the preset standard path, the potential abnormal bright spot target is judged to be a real abnormal bright spot target; when the motion trajectory characteristics of the potential abnormal bright spot target indicate that its motion speed change exceeds the normal change range in the background model database, the potential abnormal bright spot target is judged to be a real abnormal bright spot target.

[0014] Preferably, when calculating the initial risk assessment value based on the frequency of occurrence of abnormal bright spot targets, the following operations are performed: the total number of occurrences of real abnormal bright spot targets in the real-time video stream is counted; the size deviation is calculated by combining the size change characteristics of each real abnormal bright spot target with the reference size characteristics in the background model database; the initial risk assessment value is output based on the total number of occurrences and the size deviation; wherein the size deviation reflects the relative degree of change of the real abnormal bright spot target relative to the reference size characteristics.

[0015] Preferably, when using a feature matching algorithm to perform similarity grouping analysis on all real abnormal bright spot targets, the following operations are performed: the morphological contour features and size change features of each real abnormal bright spot target are combined into a feature vector representation; the feature matching algorithm is used to calculate the similarity scores between all feature vectors; a similarity score threshold is set; the feature matching algorithm is used to identify feature vector groups whose similarity scores exceed the similarity score threshold; dynamic clustering processing is performed on each feature vector group; it is determined whether there are dynamic flickering bright spot targets based on the results of the dynamic clustering processing; when the feature vector group contains at least two real abnormal bright spot targets and their morphological contour features show periodic changes, the real abnormal bright spot targets in the feature vector group are determined to be dynamic flickering bright spot targets.

[0016] Preferably, when the adjustment factor is determined based on the flickering pattern characteristics of the dynamic flickering bright spot target to correct the initial risk assessment value, the following operations are performed: the flickering frequency characteristics and flickering intensity characteristics of the dynamic flickering bright spot target are extracted; the similarity between the flickering frequency characteristics and flickering intensity characteristics and the reference flickering characteristics in the historical correction record is calculated; the adjustment factor is selected according to the similarity calculation result; when the similarity between the reference flickering characteristics in the historical correction record and the current flickering frequency characteristics and flickering intensity characteristics exceeds a preset similarity threshold, the corresponding reference adjustment factor in the historical correction record is used as the current adjustment factor; when the similarity of all reference flickering characteristics does not exceed the preset similarity threshold, the adjustment factor is calculated according to the total number of dynamic flickering bright spot targets; the initial risk assessment value is multiplied and corrected using the adjustment factor to output the corrected risk assessment value.

[0017] Preferably, when calculating the adjustment factor according to the total number of dynamic flashing bright spot targets, the following operations are performed: the size of the adjustment factor is in direct proportion to the total number of dynamic flashing bright spot targets.

[0018] Preferably, when adjusting the subsequent dynamic tracking frame rate setting based on the revised risk assessment value, the following operations are performed: obtaining the revised risk assessment value; calculating the frame rate adjustment coefficient based on the size of the revised risk assessment value; the frame rate adjustment coefficient is inversely proportional to the revised risk assessment value; and using the frame rate adjustment coefficient to multiply the current dynamic tracking frame rate to output the adjusted subsequent dynamic tracking frame rate.

[0019] Preferably, after the adjusted subsequent dynamic tracking frame rate is set, the following operations are performed: the adjusted subsequent dynamic tracking frame rate is applied to the newly acquired real-time video stream data set; the visual bright spot target is re-detected and the background model database is updated; the flickering pattern characteristic changes of the dynamic flickering bright spot target are continuously monitored; when the flickering pattern characteristic changes exceed the preset change threshold, the adjustment factor is recalculated and the risk assessment value is iteratively corrected; and the updated risk assessment value and dynamic tracking frame rate configuration are output.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] Collecting baseline video sequences to build a background model database provides a reference for subsequent analysis and processing, enabling better adaptation to background environments in different scenarios. By determining the dynamic tracking frame rate based on the size of the real-time video stream, computing resources can be rationally allocated while ensuring tracking effectiveness, avoiding unnecessary resource consumption.

[0022] The system detects visually bright targets in live video streams and calculates their average brightness as a baseline detection threshold. This allows the threshold setting to better reflect the lighting conditions of the real-time scene, reducing false or missed detections due to lighting variations. It analyzes and processes visually bright targets based on the baseline detection threshold and a background model database, combining multiple aspects of information to identify and flag potentially abnormal bright targets, improving the reliability of initial screening for abnormal targets.

[0023] When a potential abnormal bright spot target is present, its spatial position information and motion trajectory characteristics are extracted to determine whether it is a true abnormal bright spot target. This integration of the target's spatial and motion characteristics makes the identification of true abnormal targets more convincing. An initial risk assessment value is calculated based on the frequency of abnormal bright spot targets. Combined with the extracted morphological contour features and size variation characteristics, a feature matching algorithm is used to perform similarity grouping analysis, enabling more detailed classification of abnormal targets.

[0024] Based on the similarity grouping analysis results, it is determined whether there is a dynamic flickering bright spot target. If so, an adjustment factor is determined based on its flickering pattern characteristics to correct the initial risk assessment value, and the subsequent dynamic tracking frame rate is adjusted accordingly to make the risk assessment more in line with the actual situation. At the same time, the adjustment of the tracking frame rate is more targeted and can better cope with dynamically changing scenes. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a working principle diagram of the dynamic bright spot target statistics and analysis method based on visual tracking according to the present invention;

[0026] Figure 2 Flowchart for identifying potential abnormal bright spot targets;

[0027] Figure 3 Flowchart for determination of adjustment factors and revision of risk assessment values;

[0028] Figure 4 Flowchart for dynamic tracking frame rate adjustment;

[0029] Figure 5 Flowchart for iterative frame rate updates for dynamic tracking. DETAILED DESCRIPTION

[0030] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0031] See also Figure 1 The present invention provides a method for statistics and analysis of dynamic bright spot targets based on visual tracking, the method comprising:

[0032] Collect benchmark video sequences to build a background model database.

[0033] A real-time video stream dataset is obtained, and the dynamic tracking frame rate is determined according to the data size of the dataset. All visual bright spot targets in the real-time video stream are detected and the average brightness values ​​of these targets are calculated.

[0034] The calculated average brightness value is set as the detection reference threshold.

[0035] Based on the detection benchmark threshold and the constructed background model database, the visual bright spot targets in the real-time video stream are analyzed and processed, and potential abnormal bright spot targets are identified and marked.

[0036] When there are marked potential abnormal bright spot targets, the current spatial position information and motion trajectory features of each target are extracted.

[0037] Based on the extracted spatial position information and motion trajectory features, it is determined whether the potential abnormal bright spot target is a real abnormal bright spot target.

[0038] When a target is determined to be a real abnormal bright spot target, its frequency of appearance in the real-time video stream is counted and the initial risk assessment value is calculated based on this. At the same time, the morphological contour features and size change features of the real abnormal bright spot target are extracted. A feature matching algorithm is used to perform similarity grouping analysis on all identified real abnormal bright spot targets based on their morphological contour features and size change features.

[0039] Based on the results of the similarity grouping analysis, it is determined whether there are dynamic flickering bright spot targets; if it is determined that there are dynamic flickering bright spot targets, an adjustment factor is determined based on the flickering pattern characteristics of these targets, and the aforementioned initial risk assessment value is corrected using the adjustment factor; finally, based on this corrected risk assessment value, the dynamic tracking frame rate setting applied in subsequent processing is adjusted.

[0040] Example 1: See Figure 2 When determining the dynamic tracking frame rate based on the data size of a real-time video stream dataset, a preset low data volume threshold and a preset high data volume threshold are predefined in the system configuration. The value of the preset low data volume threshold is strictly lower than the preset high data volume threshold. The system processor obtains the overall data size of the current real-time video stream dataset, which typically reflects comprehensive characteristics such as the video stream duration, resolution, or total number of pixels. This data size is sequentially compared with the preset low data volume threshold and the preset high data volume threshold, generating corresponding comparison result codes. Based on the generated comparison result codes, the system automatically outputs the corresponding dynamic tracking frame rate configuration instructions. If the data size value is equal to or lower than the preset low data volume threshold, the frame rate configuration instructions specify that the system enter low frame rate mode. The frame rate value corresponding to low frame rate mode is typically within a low fixed range, such as 5 to 10 frames per second. If the data size value is greater than the preset low data volume threshold but less than or equal to the preset high data volume threshold, the frame rate configuration instructions specify that medium frame rate mode be activated. The frame rate value corresponding to medium frame rate mode is higher than the low frame rate value and falls within the intermediate range, such as 15 to 25 frames per second. If the data volume exceeds the preset high data volume threshold, the frame rate configuration command specifies switching to high frame rate mode, which corresponds to the highest frame rate, such as processing image data at 30 frames per second or higher. Once the system completes the mode setting, the video processing unit acquires and processes the image sequence at the corresponding frame rate of the selected low, medium, or high frame rate mode.

[0041] In the process of analyzing and processing visual bright spot targets based on the detection benchmark threshold and background model database, the system processor scans each frame of the real-time video stream pixel by pixel, identifies and locates all visual bright spot targets, and assigns a unique identifier to each detected target. For each identified visual bright spot target, the system extracts its average brightness value or brightness distribution statistics. The brightness statistics of the target are numerically compared and analyzed with the detection benchmark threshold previously calculated to generate a brightness deviation report. At the same time, the system extracts the key morphological features of the target, including but not limited to the target's outline shape approximation polygon, aspect ratio, area ratio, outline complexity index, and structured feature vectors such as Fourier descriptors. The morphological feature vector is sent to the background model database to perform a matching query, and the query operation traverses all pre-stored standard morphological feature records in the database. The matching check process calculates the similarity measure between the query vector and the vector stored in the database.

[0042] Combining the brightness comparison analysis report and the matching results returned by the morphological feature database query, the system executes the logic for identifying and marking potential anomalous bright spot targets. The brightness analysis module sets a specific range defined by system configuration parameters, such as a 20% range above or below the brightness threshold. When the brightness statistics of a visual bright spot target are above or below the fixed detection threshold, regardless of the morphological feature matching results, the system's decision logic module outputs an instruction to mark the target as a potential anomalous bright spot target and append a specific abnormal identification code to its unique identifier. The morphological matching check module independently executes its decision logic. If a database query for a target's morphological feature vector fails to return any valid matches—meaning that the similarity scores between all stored database features and the target's features fall below the preset matching tolerance threshold—the system's decision logic module outputs an instruction to mark the target as a potential anomalous bright spot target and assign it the same identification code, regardless of the target's brightness comparison results. Once a target is marked, subsequent processing modules focus on and operate only on targets with the specified abnormal identification code. All target raw data, brightness analysis reports, morphological matching results, and marking status are recorded in the system log.

[0043] The entire data processing process utilizes a multi-threaded architecture, with the brightness contrast analysis module and the morphological feature matching module operating in parallel. The system decision logic module receives the outputs of both modules in real time. The brightness contrast analysis module calculates the absolute and relative percentage deviation of each target's brightness value from the detection threshold. The morphological feature matching module calculates the Euclidean distance between the target's outline features and the database record, using a nearest neighbor search algorithm in a multidimensional feature space. It then returns the closest records and their similarity scores. The system decision logic module determines the marking threshold based on the configured policy. When the output of any module triggers an abnormal condition, the decision logic immediately updates the target status. The database connection management subroutine maintains a connection pool for the background model database to optimize query response time. In low frame rate mode, morphological matching may use simplified feature vectors and relax matching tolerances to improve processing efficiency. In high frame rate mode, however, the full-dimensional feature vector is used for precise matching, and strict tolerance threshold judgment is enforced. Marked target data, along with its spatial coordinates, is highlighted in the visualization interface, triggering the initiation of subsequent processing units. The analysis results also serve as input for the next stage of verifying the authenticity of potential targets. The system releases memory resources after routine data recording for normal targets that have not triggered abnormal conditions. The abnormal marking status remains in effect until the target leaves the monitoring area or is ultimately classified as a true abnormal bright spot target or the risk is eliminated.

[0044] Example 2: When determining whether a potential abnormal bright spot target is a real abnormal bright spot target based on the extracted spatial position information and motion trajectory characteristics, the system calls the spatial trajectory analysis module. The input of this module is a list of all visual bright spot targets marked as potential abnormalities in the previous step and their associated data. The data packet of each potential abnormal bright spot target contains its unique identifier, a timestamp sequence, and a set of spatial coordinates corresponding to each time point. The spatial coordinates are represented by a two-dimensional or three-dimensional coordinate system established in the monitoring scene, and the origin and scale of the coordinate system are determined by the system calibration parameters. The module preprocesses the target motion trajectory data, eliminates coordinate noise points, and performs smooth interpolation processing on the coordinate sequence to generate a continuous trajectory curve.

[0045] The preset standard path information is stored in the system path configuration library. The path configuration library contains several predefined reference motion trajectory templates. Each template trajectory is stored in the form of a set of control point coordinate sequences and interpolation functions, and is associated with a specific target category identifier. The system performs point-to-point alignment and comparison on the actual motion trajectory point sequence of each potential abnormal bright spot target with the corresponding category template trajectory in the path configuration library. The alignment process uses a temporal registration algorithm and a spatial transformation matrix to adjust the temporal and spatial coordinate system consistency of the target trajectory and the template trajectory. The system calculates the cumulative vertical distance of the target trajectory point sequence from the template trajectory. If the cumulative value exceeds the preset distance deviation threshold and continues for more than the preset number of consecutive frames, the target motion trajectory is determined to have deviated from the standard path. The preset distance deviation threshold is dynamically set according to the target size and scene ratio, and the deviation judgment result is output and recorded by the trajectory analysis submodule.

[0046] The motion trajectory feature analysis subsystem is activated. This subsystem processes the original displacement data and the time interval between consecutive frames of each potential abnormal target, and calculates its feature vectors such as average speed, instantaneous speed extreme value and speed change standard deviation. The background model database has a special partition to store the normal motion speed variation range data of different target categories, including the reasonable range of the mean speed, the maximum tolerance standard deviation of speed fluctuation, and the typical acceleration characteristic curve of a specific road section. The system compares and analyzes the motion speed feature vector calculated by the target with the normal variation range of similar targets stored in the database. The comparison operation adopts a combination of boundary checking and trend similarity evaluation: if the target speed mean exceeds the preset normal value range, the instantaneous speed extreme value exceeds the maximum speed variation range allowed for this category, or the speed variation standard deviation is higher than the preset fluctuation tolerance upper limit, then the judgment condition that the motion speed feature exceeds the normal variation range is met.

[0047] The system abnormality judgment logic controller receives independent output reports from the spatial trajectory deviation analysis submodule and the motion speed feature analysis submodule. If the judgment status report output by any submodule triggers an abnormal condition (that is, the trajectory deviates from the standard path or the speed change exceeds the normal range), the controller will update the status mark of the corresponding target identifier to a real abnormal bright spot target and generate an abnormality type code (the type code distinguishes between trajectory deviation type and speed abnormality type). The judgment result is written to the global target status table, and the data storage module is triggered to record the complete spatial position information, motion trajectory feature vector, judgment basis data and timestamp log. The target status update operation notifies the subsequent processing flow to enable the operation unit for the real abnormal bright spot target. The potential abnormality mark of the target that has not triggered any abnormal conditions is removed and restored to the normal target tracking state.

[0048] To calculate the initial risk assessment based on the frequency of occurrences of true anomalous bright spot targets, the statistical calculation module queries the global target status table during the current video stream processing cycle and filters all entries marked as true anomalous bright spot targets. The system iterates through the occurrence records of these target entries in the sequence of consecutive video frames, counting the number of occurrences of each target individually, and then calculates the total number of occurrences of all true anomalous bright spot targets within the current statistical time window. The span of this time window is set by system configuration parameters and typically covers the entire valid frame length of the currently processed video segment. This total number of occurrences serves as a fundamental input for risk assessment.

[0049] The dimensional deviation calculation engine executes in parallel. This engine accesses a list of entities currently identified as true anomalous bright spot targets and extracts the dimensional variation characteristics recorded in the feature database for each target in the list. Target dimensional variation characteristics typically include pixel area or axial length sequence information for the target region. The background model database includes a reference dimensional feature partition, which stores baseline dimensional data and historical statistics for different target categories under standard conditions. The system reads the target's category attribute identifier and, based on this, retrieves the corresponding reference dimensional feature dataset (which may include statistical mean, mode, fitted curve equation, etc.) from the reference dimensional feature partition. The calculation engine matches the target's current dimensional measurement value or sequence statistics with the reference value to generate a dimensional deviation indicator. The deviation indicator can be designed as: the absolute difference between the current measurement value and the reference mean; the percentage change of the difference as a percentage of the reference mean; or a time series-based distance metric (such as dynamic time warping distance). The deviation calculation results are recorded in a memory structure and associated with the corresponding target identifier.

[0050] The initial risk assessment generator combines two data sources: an array of the total number of occurrences of actual anomalous bright spot targets and a dataset of calculated dimensional deviations for all actual anomalous targets. The system processes this data set using a weighted algorithm model. The total number of occurrences is linearly mapped to an impact score A using a preset weighting factor. The dimensional deviations of all targets are normalized, and then their statistical average or weighted average is calculated. This average is then mapped to an impact score B using a preset conversion factor and a nonlinear correction function. Score A and score B are then combined using a rule-based model (such as a weighted summation or maximum function) to output a raw initial risk assessment value. This value is then scaled to a preset, unified assessment scale (e.g., between 0 and 10) to produce a comparable initial risk assessment result. The resulting initial risk assessment value is written to the system's risk assessment status table, along with a snapshot of key parameters involved in the calculation, including the time window definition, the number of targets, and dimensional deviation distribution statistics. This value serves as the baseline assessment variable for subsequent adjustments.

[0051] Example 3: See Figure 3In the process of implementing similarity grouping analysis on all real abnormal bright spot targets using feature matching algorithms, the system initializes the grouping analysis engine. The engine loads all instances that are currently judged to be real abnormal bright spot targets and their associated feature data. The data record of each target instance contains its unique identifier, timestamp sequence, morphological contour feature array, and size change feature matrix. The morphological contour feature array stores the target's contour polygon vertex coordinate set or contour Fourier descriptor sequence in continuous frames; the size change feature matrix records the target's main axis length, area value, and aspect ratio change curve at each time point. The grouping analysis engine constructs a composite feature vector for each target instance: the morphological contour feature array and the size change feature matrix are aligned according to the time dimension, and merged into a unified multi-dimensional feature representation vector through a feature fusion algorithm. The vector dimension is determined by the number of contour feature points and the number of size parameters, and a fixed-length digital sequence is generated using equal-weight splicing.

[0052] The system uses the configured feature matching algorithm core (a combination of improved cosine similarity algorithm and dynamic time warping algorithm), which calculates the feature vectors between all real abnormal bright spot targets. The similarity score calculation process includes the following: normalizing each vector to eliminate dimensional differences; calculating directional similarity in the vector space; and evaluating the morphological consistency of feature sequences over time. The algorithm automatically generates an N×N dimensional similarity score matrix SM, where N is the total number of current targets. Set the system preset similarity score threshold (The value range is 0.6-0.8, which is configurable). The algorithm scans the entire SM matrix and identifies all , and classify target identifiers with high similarity into the same candidate group.

[0053] The clustering execution module receives the candidate group relationship graph and applies the incremental hierarchical clustering algorithm to process it. The algorithm establishes a similarity tree structure: taking the candidate group as the initial clustering unit, iteratively merges the candidate groups with similarity exceeding The adjacent units form hierarchical clusters. During the clustering process, the distance calculation strategy is dynamically adjusted to After clustering, several final feature vector groups are output. , each A set of feature vectors containing at least two target instances. The system analysis engine Perform periodic pattern detection: extract the time series of the morphological contour features of all targets in the group, perform Fourier spectrum analysis and autocorrelation function calculation. When a significant spectrum peak is detected and the contour changes in three consecutive cycles meet the preset periodic fluctuation template, the target is judged to be a target. The corresponding target set belongs to the dynamic flashing bright spot target, and its periodic parameter indicators are recorded.

[0054] For the set of identified dynamic flickering bright spot targets, the system activates the flicker feature extraction unit. This unit calculates the key flicker pattern features from its original brightness data sequence for each target instance: flicker frequency feature (number of complete brightness cycle changes per unit time) and flicker intensity characteristics (The relative difference percentage between the highest brightness and the lowest brightness within the cycle). The feature extraction process uses a sliding window peak detection algorithm, and the window size is adaptive to the duration of the target appearance. and The composed feature pair (F, I) is submitted to the historical revision record database for similarity matching query.

[0055] The historical correction record database is designed as a multi-layer index structure: LSH local sensitive hashing is used to bucket the historical reference flash features. The similarity calculation process is defined as:

[0056]

[0057] in, Indicates the current flashing frequency The difference from the historical reference frequency, Indicates the flash intensity Difference from the reference intensity; is the frequency difference weight coefficient (typical value 0.7), is the strength difference weight coefficient (typical value 0.3); The smoothing factor is used to prevent division by zero errors (fixed value 0.01). The system calculates the current (F, I) feature and all historical reference features ( , )of value, set the preset similarity threshold . Check if the history exists. Matching items: If there is a matching record, extract the reference adjustment factor associated with the record As the current adjustment factor; if there is no matching record, the adjustment factor calculation procedure is executed: ( The total number of dynamic flashing bright spot targets, is the proportional coefficient, the preset value is 0.05-0.2 and is adjustable). Get the adjustment factor After that, the risk correction module performs the following operations: Read the initial risk assessment value , calculate the correction value ,output the correction result to the system risk assessment register.,The entire processing process generates an audit log to record the,adjustment factor calculation path and parameter trajectory.

[0058] During the feature matching process, the similarity matrix SM is generated using multi-threaded block computing technology, and GPU acceleration is automatically enabled when the number of targets N>100. The merging threshold of the clustering algorithm is adaptively adjusted according to the size of the feature vector group: a loose merging strategy is used for large-scale clusters, and strict similarity verification is implemented for small-scale clusters. The detection adopts an anti-interference design, and uses multi-resolution wavelet analysis to filter out false periodic signals caused by environmental noise. The historical correction record database implements an inert update mechanism, automatically archiving new flashing pattern characteristics as new reference records every 24 hours. The value of is limited to the system preset safety boundary [0.5,5.0] to avoid extreme correction results. After the risk correction operation is executed, the original and correction parameters The corresponding relationship is highlighted on the visual interface for operator review. The system sets an exception handler to monitor the time overhead of feature matching. In case of timeout, it automatically downgrades to a simplified matching mode to ensure real-time performance.

[0059] Example 4: See Figure 4 When the system confirms that no matching reference can be found in the historical correction record according to the rules of Example 3, the adjustment factor calculation program based on the total number of dynamic flashing bright spot targets is automatically triggered. Assume that the system detects 12 dynamic flashing bright spot target instances during the current monitoring period. Target number counter Send the integer value 12 to the adjustment factor generator. The generator presets the linear proportional coefficient k=0.15 (this parameter is stored in the system configuration file and can be adjusted dynamically). Execute the calculation logic: Adjustment factor The calculation process is completed in an independent arithmetic unit, and the result is rounded to two decimal places. The system also records the calculation log including the timestamp, input value and parameter source.

[0060] Obtaining a revised risk assessment value (This value comes from the output of the risk correction module in Example 3) After that, the system activates the frame rate adjustment control module. The preset basic constant C = 220 (reflecting the maximum processing capacity of the system). The frame rate adjustment coefficient calculation unit performs the following operation: divide the constant C by the modified risk value to obtain the adjustment coefficient This calculation is performed in the floating point unit and the result retains two significant digits. The system detects that the current dynamic tracking frame rate is in high frame rate mode, and its configuration value Frames per second. The frame rate calculation engine performs: Frames / second. Given that the physical device supports a maximum of 200 frames / second, the system automatically activates the upper threshold protection mechanism and locks the output frame rate at 200 frames / second.

[0061] The following table shows the system's processing and response logic in six typical scenarios:

[0062] Number of dynamic flashing targets Risk Modified Value Current frame rate mode Calculating the frame rate value Final execution frame rate Constraint processing type 5 6.20 Low frame rate (10) 35.48×10≈355 120 Hardware upper limit truncation 8 7.50 Medium frame rate (20) 29.33×20≈587 200 Multi-objective optimization degradation 12 8.60 High frame rate (30) 25.58×30≈767 200 Physical device limiting 3 5.80 Low frame rate (10) 37.93×10≈379 120 Energy efficiency management constraints 15 9.40 High frame rate (30) 23.40×30≈702 180 Thermal protection frequency reduction 1 4.20 Medium frame rate (20) 52.38×20≈1047 100 Communication bandwidth limitations

[0063] The process of converting the calculated frame rate value into the actual execution frame rate undergoes multiple constraint processing: the frame rate mapping controller receives the original calculated value and passes through the constraint filters of the three dimensions of physical layer, environmental layer and policy layer in turn. The physical layer filter checks the technical specifications of the camera sensor chip and discards the values ​​that exceed the maximum sampling rate of the photosensitive element. The environmental layer filter calls the real-time temperature monitoring data. When the chassis temperature exceeds 60°C, the frequency reduction algorithm is automatically activated, and the frame rate value is proportionally reduced according to the proportion of temperature exceeding the standard. The policy layer filter applies the preset resource allocation rules: when the network transmission bandwidth occupancy rate exceeds 85%, the bandwidth protection mechanism is activated, and the execution frame rate is based on the formula Dynamic compression.

[0064] The system implements a smooth transition technology when performing frame rate switching operations. Taking the above 12-target scenario as an example: the operation of switching from 30 frames / second to 200 frames / second is implemented in three stages. The first stage (0-500 milliseconds) gradually increases the acquisition rate by 30%; the second stage (500-1000 milliseconds) enables the frame buffer pool preloading mechanism to fill the data gaps; the third stage (after 1000 milliseconds) stabilizes at 200 frames / second. During the switching process, the data integrity monitoring module continues to operate, adding a timestamp check code to the video stream data packet every millisecond. Any check failure triggers the re-collection of data in that period. The target tracking algorithm uses motion trajectory prediction compensation technology during the frame rate change, and estimates the potential displacement of the target during the frame interval change through the Kalman filter algorithm to maintain the spatiotemporal continuity of target tracking.

[0065] A new processing resource allocation subsystem dynamically deploys computing resources based on the final execution frame rate. When the frame rate reaches 200 frames per second, the system automatically calls the 16 computing cores of the GPU acceleration processing node and allocates a dedicated memory channel to transmit the high-frame-rate video stream. The video decoder switches to lightweight mode and turns off unnecessary color space conversion modules. The disk write module activates a circular buffer management strategy, retaining only the full-frame-rate raw data of the last 20 seconds to avoid overloading the storage system. The system monitoring interface displays a real-time dashboard of frame rate adjustment parameters, including a dynamic histogram of the number of flashing targets, a risk value change curve, a calculation / execution frame rate comparison table, and other visual elements. The operator can manually set the k-factor experimental interval on the dashboard, and the system automatically records the differences in operating results under different parameters. The abnormal status response program generates a system optimization recommendation report when it detects that the deviation between the calculation frame rate value and the execution frame rate continues to exceed 50%, prompting you to upgrade the hardware equipment or adjust the monitoring scene parameters.

[0066] Example 5: See Figure 5After obtaining the adjusted dynamic tracking frame rate parameters generated according to the aforementioned process, the system's execution module automatically configures the video capture hardware and processing pipeline. These parameters are written to the video stream controller registers, overwriting the original frame rate settings. The controller then synchronously adjusts the image sensor clock signal frequency and data transmission channel bandwidth allocation based on the new parameters. The newly acquired real-time video stream dataset begins entering the system buffer at the updated frame rate specifications and is treated as an independent input source in subsequent processing.

[0067] The video stream processing core performs the entire target detection process on the new dataset: the pixel scanning unit re-identifies all visually bright targets using the same configured brightness threshold parameters; the spatial positioning engine assigns a three-dimensional coordinate system location identifier to each detected target; and the background model database enters a controlled update state, where the update operation integrates historical model data with the statistical characteristics of the new input frame sequence. The update process uses a layered and progressive approach: first, a sliding average calculation of the pixel grayscale values ​​is performed on static background areas; second, morphological restoration and texture reconstruction are performed on areas left behind by moving targets; finally, an adaptive background learning algorithm is activated for high-frequency change areas, calculating a pixel stability index. Pixels below the preset stability threshold are included in the background candidate set. The database version management subsystem automatically creates model snapshots before and after the update and establishes a difference log.

[0068] At the same time, the system flicker analysis thread focuses on the set of entities that have been classified as dynamic flickering bright spot targets. Each target entity is bound to an independent monitoring thread, which maintains its flicker feature time series buffer. The flicker frequency feature is tracked using real-time spectrum analysis: 1024 points of fast Fourier transform are performed per second to record the amplitude change trajectory of the dominant frequency component; the flicker intensity feature is calculated by moving the time window to calculate the peak-to-valley ratio, and the window length is dynamically expanded and contracted with the duration of the target. The feature change tracking unit generates a change rate report every second, including: frequency drift (the difference between the current frequency and the reference frequency), the intensity fluctuation coefficient σ (the standard deviation of the intensity within the window). The system preset change threshold envelope is defined by the four-dimensional parameter space: the upper limit of the frequency drift tolerance , intensity fluctuation coefficient warning value , continuous exceeding duration threshold , spatial consistency verification ratio The feature analysis results are written into the shared memory swap area for the judgment module to read.

[0069] When any dynamic flashing bright spot target meets the following conditions, the recalculation process is triggered: Condition 1: The frequency drift within three consecutive sampling periods Continue to exceed ; Condition 2, intensity fluctuation coefficient More than one mutation in a single cycle And maintain for more than two seconds; Condition three, more than The proportion of members undergoes simultaneous feature mutations. The decision logic controller activates the feature re-extraction command, which interrupts the current processing pipeline with the highest priority. The system captures the last five seconds of high-frame-rate raw brightness data of the target and recalculates its flicker frequency and intensity characteristics. The new feature dataset is marked as a derived calculation version and appended with a timestamp and spatial position checksum.

[0070] The historical correction record database enters forced query mode: the query conditions are expanded to include spatiotemporal environmental parameters, including geographic location codes, meteorological identifiers, and equipment operating condition tags. The similarity calculation uses an upgraded measurement algorithm, adding a weighting factor for the continuity of feature mutations. If there is no matching record, the adjustment factor calculator uses the latest dynamic flash target total number. Perform linear calculations and the value has been updated based on the retest results. The risk correction module obtains the current real-time risk assessment intermediate value , which is calculated based on the rolling result of the last ten minutes. and The multiplication output The frame rate adjustment factor generator responds immediately to risk value changes and is completed within 200 milliseconds. Coefficient calculation and target frame rate conversion.

[0071] The dynamic frame rate switching process utilizes a three-step buffering mechanism: in the first stage, a temporary acquisition rate of 70% of the target frame rate is configured; in the second stage, data pipeline preloading is enabled, filling the frame buffer to 80% of capacity; and in the final stage, the system switches to the full target frame rate. During the entire switching process, the target tracking algorithm maintains dual-track processing: existing trajectories continue to be predicted using Kalman filtering, while alternative trajectories are generated using newly acquired data. After the switch is complete, the trajectory fusion controller performs trajectory matching and merging. Trajectory segments with merging errors exceeding the pixel tolerance trigger spatial interpolation compensation. The system console displays the updated parameter matrix in real time, including key metrics such as the number of recalculation event triggers, the current feature mutation index, and the evolution curve of the iterative risk assessment value. The configuration management subsystem automatically generates a configuration change audit report every five minutes, documenting the complete state transition path and constraint validation status. All recalculation operations are marked as system optimization events, and event logs are synchronously transmitted to the remote analysis platform.

[0072] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0073] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for counting and analyzing dynamic bright spot targets based on visual tracking, comprising the following steps: Collect benchmark video sequences and build a background model database; Acquire a real-time video stream data set, determine a dynamic tracking frame rate based on the data size of the real-time video stream data set, detect all visual bright spot targets in the real-time video stream and calculate the average brightness value of the visual bright spot targets; set the average brightness value as a detection reference threshold; Analyze and process visual bright spot targets based on the detection reference threshold and background model database, and identify and mark potential abnormal bright spot targets; When there are potential abnormal bright spot targets, extract the spatial position information and motion trajectory characteristics of each potential abnormal bright spot target; and determine whether the potential abnormal bright spot target is a real abnormal bright spot target based on the spatial position information and motion trajectory characteristics; When it is determined to be a real abnormal bright spot target, the initial risk assessment value is calculated based on the frequency of the abnormal bright spot target; Extract the morphological contour features and size change features of real abnormal bright spot targets; Use feature matching algorithm to perform similarity grouping analysis on all real abnormal bright spot targets; Determine whether there is a dynamic flickering bright spot target based on the results of the similarity grouping analysis; when it is determined that there is a dynamic flickering bright spot target, determine an adjustment factor based on the flickering pattern characteristics of the dynamic flickering bright spot target to correct the initial risk assessment value; and adjust the subsequent dynamic tracking frame rate setting based on the corrected risk assessment value.

2. The method for counting and analyzing dynamic bright spot targets based on visual tracking according to claim 1, wherein when determining the dynamic tracking frame rate according to the data volume of the real-time video stream data set, the following operations are performed: comparing the data volume with a preset low data volume threshold and a preset high data volume threshold respectively; and outputting a corresponding dynamic tracking frame rate configuration according to the comparison results; wherein, The preset low data volume threshold is lower than the preset high data volume threshold; when the data volume does not exceed the preset low data volume threshold, the dynamic tracking frame rate is set to a low frame rate mode; when the data volume exceeds the preset low data volume threshold but does not exceed the preset high data volume threshold, the dynamic tracking frame rate is set to a medium frame rate mode; when the data volume exceeds the preset high data volume threshold, the dynamic tracking frame rate is set to a high frame rate mode; the frame rate value corresponding to the low frame rate mode is lower than the medium frame rate mode, and the frame rate value corresponding to the medium frame rate mode is lower than the high frame rate mode.

3. The dynamic bright spot target statistics and analysis method based on visual tracking according to claim 2, when analyzing and processing the visual bright spot targets according to the detection benchmark threshold and the background model database, performs the following operations: comparing the brightness values ​​of all visual bright spot targets in the real-time video stream with the detection benchmark threshold; and performing a matching check on the morphological features of each visual bright spot target with the standard morphological features in the background model database; identifying potential abnormal bright spot targets and marking them based on the results of the comparison and matching checks; when the brightness value of the visual bright spot target exceeds the ratio range set by the detection benchmark threshold, the visual bright spot target is determined to be a potential abnormal bright spot target and marked; when the morphological features of the visual bright spot target do not find a matching record in the background model database, the visual bright spot target is determined to be a potential abnormal bright spot target and marked.

4. According to the dynamic bright spot target statistics and analysis method based on visual tracking in claim 3, when judging whether a potential abnormal bright spot target is a real abnormal bright spot target based on the spatial position information and motion trajectory characteristics, the following operations are performed: when the spatial position information of the potential abnormal bright spot target shows that its motion trajectory deviates from a preset standard path, the potential abnormal bright spot target is judged to be a real abnormal bright spot target; when the motion trajectory characteristics of the potential abnormal bright spot target indicate that its motion speed change exceeds the normal change range in the background model database, the potential abnormal bright spot target is judged to be a real abnormal bright spot target.

5. The method for counting and analyzing dynamic bright spot targets based on visual tracking according to claim 4, wherein when calculating the initial risk assessment value based on the frequency of occurrence of abnormal bright spot targets, the following operations are performed: counting the total number of occurrences of real abnormal bright spot targets in the real-time video stream; The size deviation is calculated by combining the size change characteristics of each real abnormal bright spot target with the reference size characteristics in the background model database; An initial risk assessment value is output based on the total number of occurrences and the size deviation; wherein the size deviation reflects the relative degree of change of the actual abnormal bright spot target relative to the reference size feature.

6. According to the dynamic bright spot target statistics and analysis method based on visual tracking according to claim 5, when using the feature matching algorithm to perform similarity grouping analysis on all real abnormal bright spot targets, the following operations are performed: the morphological contour features and size change features of each real abnormal bright spot target are combined into a feature vector representation; the similarity score between all feature vectors is calculated using the feature matching algorithm; a similarity score threshold is set; the feature vector group whose similarity score exceeds the similarity score threshold is identified by the feature matching algorithm; each feature vector group is dynamically clustered; and it is determined whether there is a dynamic flickering bright spot target based on the result of the dynamic clustering process; when the feature vector group contains at least two real abnormal bright spot targets and their morphological contour features show periodic changes, the real abnormal bright spot targets in the feature vector group are determined to be dynamic flickering bright spot targets.

7. According to the dynamic bright spot target statistics and analysis method based on visual tracking according to claim 6, when determining the adjustment factor based on the flickering pattern characteristics of the dynamic flickering bright spot target to correct the initial risk assessment value, the following operations are performed: extract the flickering frequency characteristics and flickering intensity characteristics of the dynamic flickering bright spot target; calculate the similarity between the flickering frequency characteristics and flickering intensity characteristics and the reference flickering characteristics in the historical correction record; select the adjustment factor according to the similarity calculation result; when the similarity between the reference flickering characteristics in the historical correction record and the current flickering frequency characteristics and flickering intensity characteristics exceeds the preset similarity threshold, use the corresponding reference adjustment factor in the historical correction record as the current adjustment factor; when the similarity of all reference flickering characteristics does not exceed the preset similarity threshold, calculate the adjustment factor according to the total number of dynamic flickering bright spot targets; use the adjustment factor to multiply the initial risk assessment value to output the corrected risk assessment value.

8. According to the dynamic bright spot target statistics and analysis method based on visual tracking in claim 7, when calculating the adjustment factor based on the total number of dynamic flashing bright spot targets, the following operations are performed: the size of the adjustment factor is directly proportional to the total number of dynamic flashing bright spot targets.

9. The method for dynamic bright spot target statistics and analysis based on visual tracking according to claim 8, wherein when adjusting the subsequent dynamic tracking frame rate setting based on the revised risk assessment value, the following operations are performed: obtaining the revised risk assessment value; Calculate the frame rate adjustment coefficient according to the size of the revised risk assessment value; The frame rate adjustment factor is inversely proportional to the revised risk assessment value; The frame rate adjustment factor is used to multiply the current motion tracking frame rate to output the adjusted subsequent motion tracking frame rate.

10. The method for dynamic bright spot target statistics and analysis based on visual tracking according to claim 9, wherein after the adjusted subsequent dynamic tracking frame rate is set, the following operations are performed: applying the adjusted subsequent dynamic tracking frame rate to a newly acquired real-time video stream data set; Re-detect visual bright spot targets and update the background model database; Continuously monitor the changes in the flashing pattern characteristics of dynamic flashing bright spot targets; When the flicker pattern characteristic changes exceed a preset change threshold, the adjustment factor is recalculated and the risk assessment value is iteratively corrected; Output the updated risk assessment value and dynamic tracking frame rate configuration.

Images