Infrared time sequence verification basketball goal anti-misjudgment detection method

Abstract

The present application relates to the technical field of basketball referee, discloses an infrared time sequence verification basketball goal anti-misjudgment detection method. The method collects the infrared signal sequence of the shooting action in real time through the infrared sensor array around the basketball hoop, segments the shooting trajectory segment and extracts the space-time coupling feature, and then generates a multi-scale trajectory analysis grid, wherein the grid accuracy of the region with sharp infrared signal change is higher than that of the gentle region. The grid time sequence is analyzed for interference characteristics to generate an infrared signal interference index, the dynamic path verification algorithm is used to calculate the abnormal deviation degree of the basketball trajectory combined with the index, and finally the misjudgment verification conclusion of the shooting action is generated according to the comparison result of the abnormal deviation degree and the preset threshold. The method can comprehensively capture the basketball trajectory, effectively eliminate the interference, improve the accuracy and real-time performance of the goal judgment, and help to improve the fairness of the basketball game judgment.

Description

Technical Field

[0001] This invention relates to the field of basketball refereeing technology, specifically to an infrared timing verification method for detecting basketball goals to prevent misjudgment. Background Technology

[0002] In basketball, the accuracy of goal decisions directly impacts the fairness and entertainment value of the game. Currently, goal decisions in basketball primarily rely on the referee's visual inspection and traditional mechanical sensors; however, these methods have significant limitations.

[0003] Visual observation is heavily influenced by the referee's perspective, reaction speed, and subjective judgment, making it prone to misjudgments in high-speed game scenarios. For example, when a basketball rapidly passes the edge of the rim or makes complex collisions with it, the referee struggles to accurately determine whether the ball has completely passed through the rim. Traditional mechanical sensing devices often use a single sensor, only detecting whether the basketball touches the rim or passes through a specific point, failing to comprehensively record the basketball's trajectory. When the basketball undergoes complex movements such as bouncing or spinning, these devices frequently produce misjudgments due to incomplete signal acquisition.

[0004] As basketball becomes increasingly professional, the demands for accurate officiating are rising. While some existing technologies utilize video replay to assist in judging, this relies on manual analysis, is time-consuming, and is susceptible to errors due to factors such as camera angle and lighting. Furthermore, the large volume of video data processing makes real-time judging difficult and fails to meet the demands of smooth gameplay.

[0005] Some sensor-based detection methods, due to improper sensor deployment or simplistic signal processing algorithms, fail to fully utilize infrared signals. When the basketball's trajectory is complex or external interference is present, these methods cannot effectively distinguish between valid and interfering signals, further increasing the possibility of misjudgments. Therefore, a goal detection method is needed that can accurately acquire the basketball's trajectory in real time and effectively eliminate interference to improve the accuracy and fairness of basketball game officiating. Summary of the Invention

[0006] The purpose of this invention is to provide an infrared time-series verification method for detecting basketball goals against false positives, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides an infrared time-series verification method for detecting basketball goals against false positives, the method comprising:

[0008] The infrared signal sequence of the shooting action is collected in real time by an infrared sensor array deployed around the basketball hoop; the shooting trajectory segment containing the shooting trajectory is segmented from the infrared signal sequence.

[0009] Spatiotemporal coupling features are extracted from the shooting trajectory segment; based on the spatiotemporal coupling features, a multi-scale trajectory analysis grid is generated according to the morphological features of the basketball trajectory; wherein, the grid accuracy corresponding to the trajectory region with drastic changes in infrared signal is higher than that of the region with gentle changes;

[0010] Interference characteristics are analyzed on the time series of multi-scale trajectory analysis grids to generate an infrared signal interference index.

[0011] Based on the infrared signal interference index, a dynamic path verification algorithm is used to calculate the abnormal deviation of the basketball trajectory; based on the comparison between the abnormal deviation and the preset threshold, a misjudgment verification conclusion of the shooting action is generated.

[0012] Preferably, segmenting the shooting trajectory segment containing the shooting trajectory from the infrared signal sequence includes:

[0013] Identify the locations of all trough points in the infrared signal sequence; divide the infrared signal sequence into initial trajectory subsequences based on adjacent trough points;

[0014] Calculate the signal intensity range of each initial trajectory subsequence; perform cluster analysis based on the signal intensity range of all initial trajectory subsequences to separate the effective shooting trajectory sequence;

[0015] The output of the effective shooting trajectory sequence is connected to the spatiotemporal coupling feature extraction process.

[0016] Preferably, the step of generating a multi-scale trajectory analysis mesh based on the morphological characteristics of the basketball trajectory includes:

[0017] Based on the spatial structure of the basketball hoop, the infrared sensor monitoring area is divided into multiple sub-regions for analysis.

[0018] For each sub-region of analysis, the infrared signal sampling frequency is used to determine the grid cell size and grid density of that sub-region.

[0019] Based on the aforementioned spatiotemporal coupling characteristics, independent grids are generated for each analysis sub-region.

[0020] Mesh stitching is achieved by matching the mesh boundary nodes of adjacent analysis sub-regions using unique codes;

[0021] Connect the output of the spliced ​​mesh to the interference feature analysis process.

[0022] Preferably, the generated infrared signal interference index includes:

[0023] Extract the spatiotemporal variation parameters of each grid node in the multi-scale trajectory analysis grid;

[0024] The local anti-interference fluctuation index is calculated based on the fluctuation amplitude and correlation of spatiotemporal variation parameters;

[0025] A global interference baseline is generated based on the local anti-interference fluctuation index of all grid nodes;

[0026] Connect the output of the global interference baseline to the anomaly deviation calculation process.

[0027] Preferably, the calculation of the abnormal deviation of the basketball trajectory using the dynamic path verification algorithm includes:

[0028] Construct a standard motion path model for basketball trajectory;

[0029] Dynamically correct the standard motion path model based on the infrared signal interference index;

[0030] The instantaneous deviation is calculated by the deviation between the real-time trajectory coordinates and the corrected path model;

[0031] The instantaneous deviation of the entire shooting process is used to generate a comprehensive abnormal deviation.

[0032] Preferably, the method further includes:

[0033] Establish a confidence label library for shooting motions based on historical shooting data;

[0034] When the overall abnormal deviation exceeds the preset threshold, the shooting action confidence label library is called for secondary verification.

[0035] The secondary verification results are fed back into the misjudgment verification conclusion generation process.

[0036] Preferably, the update of the shooting motion confidence label library based on historical shooting data includes:

[0037] Extract the misjudgment type features from the misjudgment verification conclusion;

[0038] Adjust the weighting coefficient of the shooting motion confidence label based on the characteristics of the misjudgment type;

[0039] The updated weighting coefficients are then input into the secondary verification process.

[0040] Preferably, the method further includes a dynamic resource allocation mechanism:

[0041] Real-time monitoring of the infrared signal processing load of each analysis sub-region;

[0042] Based on the mapping relationship between signal processing load and grid computing time, grid computing resources are dynamically allocated to high-load sub-regions;

[0043] The resource allocation results are fed back to the independent grid generation process.

[0044] Preferably, the execution of the dynamic resource allocation mechanism includes:

[0045] When the signal processing load of a certain analysis sub-region exceeds the warning threshold, it is marked as a critical monitoring area;

[0046] Reduce the grid precision in non-critical monitoring areas to free up computing resources;

[0047] The freed-up computing resources will be reallocated to the grid for high-precision processing of critical monitoring areas.

[0048] Preferably, the method further includes:

[0049] Statistical analysis of the probability of misjudgments in consecutive shots in key monitoring areas;

[0050] When the false positive probability continues to exceed the risk threshold, the full-area high-precision grid verification mode is triggered.

[0051] Connect the output of the full-area high-precision grid to the interference feature analysis process.

[0052] Compared with the prior art, the beneficial effects of the present invention are:

[0053] By deploying an array of infrared sensors around the basketball hoop to collect the infrared signal sequence of the shooting motion in real time, the movement of the basketball around the hoop can be captured comprehensively and meticulously. Compared with traditional single sensors, the array-deployed sensors can acquire signals from multiple angles, reducing blind spots in signal acquisition and making subsequent trajectory analysis more complete.

[0054] By segmenting the shooting trajectory segment containing the shooting trajectory from the infrared signal sequence, the effective signal can be accurately extracted. This process eliminates redundant signals unrelated to the shooting trajectory, reduces the complexity of subsequent data processing, makes the analysis more targeted, and helps improve the efficiency of trajectory analysis.

[0055] Spatiotemporal coupling features are extracted and a multi-scale trajectory analysis mesh is generated, fully considering the dynamic changes in the basketball trajectory. The mesh precision is adjusted according to the intensity of infrared signal changes; a higher precision mesh is used in complex signal regions to more clearly present trajectory details, while the mesh precision is reduced in regions with flat signals to avoid unnecessary waste of computational resources. This adaptive meshing method allows trajectory analysis to balance detail and efficiency, improving the ability to analyze the morphological characteristics of the basketball trajectory.

[0056] Interference characteristic analysis of time series data from multi-scale trajectory analysis grids and the generation of an infrared signal interference index can effectively identify and quantify the impact of external interference on signals. This index distinguishes between valid signals generated by basketball motion and interference signals from the external environment, reducing the influence of interference factors on the judgment results and providing a more reliable basis for subsequent trajectory verification.

[0057] Based on the infrared signal interference index, a dynamic path verification algorithm is used to calculate the abnormal deviation of the basketball trajectory, enabling dynamic tracking of the basketball's movement path. This algorithm combines the interference index with real-time adjustments to the verification strategy. When interference is present, the algorithm calculates the degree of trajectory deviation to determine whether the basketball's movement conforms to a normal shooting trajectory. This dynamic verification method enhances the ability to judge complex motion trajectories, accurately identifying the movement trend even when the basketball undergoes complex movements such as bouncing and spinning.

[0058] The system generates a false positive verification conclusion based on the comparison between the abnormal deviation and a preset threshold, creating a complete closed loop in the detection process. By using explicit threshold judgments, it avoids subjective assumptions inherent in manual judging, making the results more objective and consistent. Furthermore, the entire process is based on real-time acquisition of infrared signals for processing and analysis, enabling rapid generation of verification conclusions to meet the real-time judging requirements of basketball games and ensuring the smooth flow of the game. Attached Figure Description

[0059] Figure 1 This is a schematic diagram illustrating the working principle of the infrared timing verification method for detecting basketball goals against false positives as described in this invention.

[0060] Figure 2 A flowchart for segmenting shooting trajectory segments;

[0061] Figure 3 A flowchart for calculating abnormal deviation;

[0062] Figure 4 Flowchart for updating the confidence label library;

[0063] Figure 5 A flowchart for resource allocation in key monitoring areas. Detailed Implementation

[0064] 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 are only some embodiments of the present invention, and not all embodiments. 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.

[0065] Please see Figure 1 This invention provides an infrared time-series verification method for detecting basketball goals against false positives, the method comprising:

[0066] An infrared sensor array deployed around the perimeter of the basketball hoop continuously collects infrared signal sequences of the shooting motion. The infrared signal sequences are segmented to extract shooting trajectory segments containing valid shooting trajectories. Spatiotemporal coupling features are extracted from these trajectory segments. Based on these spatiotemporal coupling features, a multi-scale trajectory analysis grid is generated according to the morphological characteristics of the basketball's trajectory; regions with drastic changes in the infrared signal correspond to higher-precision grid divisions, while regions with gradual changes correspond to lower-precision grids. Interference feature analysis is performed on the time-series data of the multi-scale trajectory analysis grid to generate an infrared signal interference index. Based on the infrared signal interference index, a dynamic path verification algorithm is used to calculate the abnormal deviation of the basketball trajectory. Based on the comparison between the abnormal deviation and a preset threshold, a false positive verification conclusion for the shooting motion is output.

[0067] Example 1: See Figure 2 An infrared sensor array continuously collects infrared signal sequences around the perimeter of the basketball hoop. The sequence segmentation process first locates all trough points, determined by detecting local minimum signal strength and their time derivative characteristics. Signal segments between adjacent trough points are automatically divided into initial trajectory subsequences. For each initial trajectory subsequence, the signal strength range is calculated; this range is the absolute difference between the maximum and minimum signal strength values ​​within that subsequence. The range values ​​of all initial trajectory subsequences are input into a clustering analysis module. The clustering analysis employs an unsupervised learning mechanism, automatically defining cluster boundaries based on the distribution density of the range values. High-density range clusters close to the cluster centers are identified as valid shooting trajectory sequences. These valid shooting trajectory sequences retain their original timestamps and spatial location information and are output to the spatiotemporal coupling feature extraction module.

[0068] The physical structure of the basketball hoop divides the monitoring area into the area directly above the rim, the inner area of ​​the rim, the front area of ​​the backboard, and the support connection area. Each sub-region is assigned grid parameters based on its preset infrared signal sampling frequency: the area directly above the rim (sampling frequency F1) uses the smallest grid cell size S1 and the highest grid density D1; the front area of ​​the backboard (sampling frequency F2) uses a medium grid cell size S2 and a grid density D2; and the support area (sampling frequency F3) uses the largest grid cell size S3 and the lowest grid density D3. After the spatiotemporal coupling characteristics are input into the grid generation module, each sub-region generates an independent grid structure in parallel. The nodes of each independent grid contain three-dimensional spatial coordinates and timestamp identifiers.

[0069] Adjacent sub-region grid boundary nodes are matched using a unique encoding rule: adjacent boundary nodes between the area directly above the rim and the area inside the rim are assigned the prefix "BOUNDARY_A", and adjacent boundary nodes between the area in front of the backboard and the support area are assigned the prefix "BOUNDARY_B". Encoding matching is achieved through a hash mapping table; when the spatial coordinate difference between adjacent nodes is less than a preset tolerance threshold... And the timestamp difference is less than the synchronization threshold. Upon successful matching of boundary nodes, the system determines that the boundary nodes have been matched. Successfully matched nodes establish bidirectional topological connections, forming a seamless composite mesh structure. During mesh stitching, the spatial coordinates of the boundary nodes are automatically corrected to maintain mathematical continuity between adjacent mesh edges. In the final multi-scale trajectory analysis mesh, the mesh accuracy reaches the highest level in the region directly above the basket where infrared signals change drastically, while the mesh accuracy in the support region where signal changes are gradual remains at a basic level. The stitched mesh object is output to the interference feature analysis module, simultaneously recording the mesh accuracy level and boundary topological relationships of each sub-region.

[0070] The timing logic of signal sequence segmentation and mesh generation strictly corresponds: once a valid shooting trajectory sequence is identified, its timestamp range automatically activates the corresponding time period's mesh analysis region. The system dynamically loads the mesh generation parameters for the corresponding sub-region based on the spatial distribution coordinates of the shooting trajectory. In continuous shooting scenarios, the mesh parameters for each sub-region are dynamically refreshed according to the real-time signal sampling frequency; if the sampling frequency changes beyond a threshold... A real-time grid density adjustment mechanism is triggered. The grid cell size is adjusted using a progressive scaling strategy to avoid discontinuities in trajectory analysis due to abrupt parameter changes. A time-aligned mapping is established between the spatial coordinate data of all grid nodes and the original infrared signal intensity values, forming a four-dimensional grid data structure (x, y, z, t) with a time dimension. This data structure allows subsequent modules to extract spatiotemporal variation parameters of grid nodes by time slice.

[0071] Example 2: See Figure 3 Each grid node in the multi-scale trajectory analysis grid contains a preset number of spatiotemporal variation parameters. The system extracts the signal intensity change rate parameter in the time dimension, which represents the change in infrared signal intensity per unit time; and the position offset parameter in the spatial dimension, which records the displacement of the node's spatial coordinates in adjacent time slices. The spatial position offset of the grid node is calculated using the Euclidean distance formula in a three-dimensional coordinate system, with the time slice interval fixed at Δt milliseconds. The signal intensity change rate is obtained by dividing the intensity difference of the node in consecutive time slices by the time interval.

[0072] The calculation of the local anti-interference fluctuation index is based on two dimensions: the fluctuation amplitude of spatiotemporal variation parameters between adjacent nodes, and the correlation between different parameters. The fluctuation amplitude is obtained by comparing the parameter differences between the current node and its neighboring nodes, and the correlation analysis uses the parameter covariance matrix. For each grid node, the system constructs a local evaluation window containing its directly adjacent nodes. The spatiotemporal variation parameters of the nodes within the window are input into the index generation module, which calculates the average parameter difference between the central node and surrounding nodes in the window, and simultaneously analyzes the correlation coefficient between the signal strength change rate and the position offset. The difference and correlation coefficient are weighted and fused to generate the local anti-interference fluctuation index for that grid node. A higher index value indicates weaker signal stability in that area.

[0073] The local anti-interference fluctuation index of all grid nodes is transmitted to the aggregation module. The aggregation module adopts a hierarchical aggregation strategy: first, physically adjacent nodes are divided into multiple local blocks, and the arithmetic mean of the index within each block is calculated; then, the block average is input into the global calculation unit, and instantaneous fluctuation noise is eliminated through a time sliding window mechanism. The global interference baseline output is a continuous curve that changes over time, and its value reflects the overall signal stability status of the entire monitoring area. This baseline is transmitted in real time to the input interface of the dynamic path verification algorithm.

[0074] The standard trajectory model of a basketball is stored in a database, containing a parametric expression of the ideal parabolic equation. The core parameters of the model include the initial velocity of the shot, the shooting angle, the gravitational acceleration constant, and the air resistance coefficient. The model pre-defines physical constraints such as the basketball diameter, the height of the basketball hoop, and the distance to the backboard. Based on the real-time received infrared signal interference index, the system dynamically corrects the standard trajectory model: when the interference index exceeds the correction threshold α, the model automatically increases the air resistance compensation coefficient; when the interference index is below the stability threshold β, it operates using the basic parameters. The correction parameters are implemented through a lookup table mechanism; the system has a built-in mapping table between the interference index and the correction coefficient.

[0075] Real-time trajectory coordinates are obtained through spatial position interpolation of multi-scale grid nodes. Within each time slice, the system compares the interpolated coordinates with the theoretical coordinates of the corrected path model. Instantaneous deviation calculation includes spatial deviation assessment and temporal synchronization calibration: spatial deviation is calculated using the Euclidean distance formula in three-dimensional space; temporal synchronization calibration is achieved by aligning the trajectory coordinate timestamps with the phase of the model's time axis. The instantaneous deviation for each time slice is written to a circular buffer, the depth of which covers the entire time period of a single shot.

[0076] The overall abnormal deviation is automatically calculated at the end of the shooting motion. The system reads all instantaneous deviation data in the circular buffer and processes it using a time-weighted integral algorithm: the instantaneous deviation of the early trajectory segment is assigned a lower weight coefficient, while the critical trajectory segment near the basket is assigned a higher weight coefficient. The weight coefficient distribution curve is preset according to the shooting kinematics characteristics. A time decay factor is also introduced during the integration process to avoid the early deviation value from having an excessive impact on the result. The final output of the overall abnormal deviation is a normalized scalar value, whose numerical range is mapped to the preset evaluation range. This value is directly input into the judgment logic of the misjudgment verification conclusion generation module.

[0077] The data processing workflow establishes a strict time constraint mechanism: the time delay from grid node parameter extraction to instantaneous deviation calculation does not exceed T milliseconds; the update frequency of the global interference baseline is synchronized with the refresh cycle of the multi-scale grid. All intermediate calculation results are appended with timestamps and spatial region identifiers to support retrospective analysis in subsequent modules. When the system detects a sudden change in instantaneous deviation within a continuous time slice, it automatically triggers a secondary sampling verification process for trajectory coordinates to eliminate miscalculations caused by transient sensor interference.

[0078] Example 3: See Figure 4 Historical shooting data is stored in a distributed database cluster, including timestamps of shooting events, three-dimensional trajectory coordinate sequences, infrared signal intensity distributions, and manual review conclusion labels. The system processes raw records through a data cleaning module, extracting effective feature vectors to construct a shooting action confidence label library. The label library employs a multi-layered index structure: the first layer divides spatial regions based on the coordinates of the shooting start position; the second layer categorizes data according to shooting release speed; and the third layer associates environmental parameters, including light intensity and temperature / humidity sensor readings. Each label record contains a standard trajectory pattern feature vector, a typical interference pattern feature vector, and initial weight coefficients. The initial value of the weight coefficients is determined by the ratio of the number of samples for that label to the total number of samples.

[0079] When the overall anomaly deviation exceeds a preset threshold θ, the system activates a tag library query command. The query process first locates the three-dimensional trajectory features of the current shooting event: the spatial coordinates of the point of maximum trajectory curvature are calculated as the spatial index key, the initial shooting velocity value is used as the velocity index key, and real-time environmental sensor data is used as the environmental index key. The corresponding tag set is retrieved using multi-key combinations, returning the K most matching candidate tags. The feature matching degree is calculated using a trajectory pattern similarity algorithm, which compares the difference between the current shooting trajectory and the feature vector recorded by the tags in the temporal and spatial domains. The matching degree calculation formula is as follows:

[0080]

[0081] Where: S represents the overall difference degree, and N is the total number of trajectory sampling points. This represents the time deviation (in milliseconds) at the i-th sampling point. S represents the spatial Euclidean distance deviation (in millimeters) for the i-th sampling point, α is the time deviation weighting factor, and β is the spatial deviation weighting factor. A smaller S value indicates a higher degree of matching. The system selects the top M labels with the smallest S values ​​to generate a secondary validation result set, which includes the interference pattern analysis conclusions and confidence scores for each label.

[0082] The secondary verification result is fed back to the misjudgment verification conclusion generation module. This module performs logical arbitration between the initial anomaly deviation conclusion and the tag library conclusion: when the consistency between the two exceeds the arbitration threshold γ, the initial conclusion is output directly; when there is a conflict, the tag library conclusion overrides the initial conclusion. The arbitration result is appended with an arbitration flag and stored in the event log, simultaneously triggering the tag library update process.

[0083] The false positive type features in the false positive verification conclusion are extracted using a feature parsing engine. The engine identifies three core features: interference source category features, including electromagnetic interference identifiers and ambient light change identifiers; trajectory distortion morphology features, including trajectory break markers and abnormal velocity change points; and sensor failure features, recording the ID of the abnormal sensor node and the failure time interval. The system builds an independent-dimensional statistical histogram for each feature, and the histogram is updated on a rolling basis according to a time window.

[0084] The weighting coefficients are adjusted using an incremental update strategy. For each label involved in a current misclassified event, the system calculates the difference between its misclassification type features and historical records. :

[0085] when When σ < 0.01, the weight coefficient adjustment step size k = 0.01

[0086] When σ≤ When <2σ, k=0.03

[0087] when When ≥2σ, k=0.05

[0088] Where σ is the baseline value of feature difference, determined by the standard deviation of historical data. The weight update formula is:

[0089]

[0090] in: This represents the updated weight coefficients. The weighting coefficients before the update. To dynamically adjust the step size, Misjudgment type feature difference degree.

[0091] The upper limit constraint for the weight coefficient is 1.0, and the lower limit constraint is 0.2. Update operations employ an optimistic locking mechanism in the distributed storage layer to avoid concurrent conflicts. New weight values ​​are synchronized to the online query service in real time, and a version snapshot is generated and stored in the audit log.

[0092] The tag library maintenance background process periodically performs a global weight rebalancing: every 24 hours, it scans the weight distribution and performs a decay operation on tags that have not been updated for more than 72 hours, with a fixed decay coefficient of 0.95. During the rebalancing process, redundant tags with similarity exceeding the merging threshold η are automatically merged, releasing storage resources. All update operations maintain version rollback capability; the system retains a complete version history for the most recent 30 days and supports restoring the tag library state of a specific version by specifying a timestamp.

[0093] A data closed-loop process establishes a feedback channel: After each weight update, the system automatically retrieves secondary verification events using the label within the last 7 days and recalculates the arbitration result. When the new weight causes a change in the arbitration conclusion, a conclusion correction report is generated and pushed to the administrator console. Key trajectory data involved in the conclusion correction is automatically transferred to a high-priority storage area for subsequent in-depth analysis. A timestamp alignment mechanism ensures that all operations are performed under a unified time base, with system clock synchronization accuracy controlled within ±1 millisecond.

[0094] Example 4: See Figure 5 The system monitors the signal processing load of each analysis sub-region around the basketball hoop in real time. The load data includes three core indicators: CPU utilization, memory consumption, and processing latency. The monitoring period is fixed at 200 milliseconds, and a snapshot of the load status is generated each time. The mapping relationship between load data and grid computation time is maintained through a pre-established configuration table, which records the grid generation time baseline value corresponding to different load levels.

[0095] Table 1: Resource configuration strategies under typical load conditions.

[0096]

[0097] When the CPU utilization rate of a sub-region (e.g., D4) exceeds 85% for three consecutive monitoring periods, or the processing latency remains higher than 80 milliseconds, the system determines that the region has met the resource reallocation trigger condition. The resource scheduler immediately executes a dynamic allocation strategy:

[0098] To reduce the mesh accuracy level in non-critical areas (such as C3), specific measures include increasing the mesh cell size from the baseline of 5mm to 8mm and reducing the number of mesh boundary nodes by 40%.

[0099] The freed computing resources (including 0.8 CPU cores and 180MB of memory) were reallocated to the critical region D4.

[0100] The mesh density in key areas has been increased to the highest level, the mesh cell size has been reduced to 3mm, and the number of boundary nodes has been increased by 50%.

[0101] Resource allocation results are fed back to the mesh generation module in real time via a message queue. Upon receiving the allocation command, the module initiates a high-precision mesh generation process for key areas.

[0102] The grid cell topology has been upgraded from a basic hexahedron to an icosahedron;

[0103] The time sampling interval has been reduced from 20 milliseconds to 12 milliseconds;

[0104] The spatial coordinates of nodes are calculated using double-precision floating-point arithmetic.

[0105] Meanwhile, a simplified grid mode is enabled in non-critical areas:

[0106] Mesh cells degenerate into tetrahedral structures;

[0107] The time sampling interval has been increased to 30 milliseconds;

[0108] The node spatial coordinates are calculated using single-precision floating-point operations;

[0109] The load balancer controller continuously tracks the adjustment effects. When the CPU utilization in critical region D4 falls below 75% and the processing latency is less than 50 milliseconds, the system automatically removes the critical status flag. After removal, resource reclamation is performed.

[0110] Grid density in key areas is gradually reduced to the standard level;

[0111] The released 0.6 CPU cores and 150MB of memory are returned to the resource pool;

[0112] The mesh accuracy in non-critical areas is restored to the initial configuration;

[0113] An exception handling mechanism is activated during resource allocation: when the load in a certain area suddenly drops to the idle threshold (CPU < 30%, latency < 20ms), the idle resources in that area are immediately added to the global allocation pool. When multiple areas reach a critical state simultaneously, the system allocates resources according to the area priority strategy: the area directly above the rim is set to the highest priority, the area in front of the backboard is medium priority, and the support area is the lowest priority.

[0114] Historical load data is stored in a circular buffer, retaining the most recent 300 monitoring records. Before each resource allocation decision, the system retrieves historical load patterns: if the similarity between the current load curve and a historical anomaly pattern exceeds a matching threshold, the regular adjustment process is skipped, and the historically optimal resource configuration scheme is directly adopted. All resource allocation operations are logged in detail, including timestamps, region identifiers, resource change amounts, and adjustment durations.

[0115] Example 5: The system continuously tracks the statistical values ​​of misjudgment probability in key monitoring areas. After each shooting event is processed, the system automatically records the status of the misjudgment verification conclusion in the key area. The misjudgment probability is calculated using a sliding time window mechanism: taking the most recent T consecutive shooting events as the statistical sample, the frequency of positive misjudgment verification conclusions is calculated, and the ratio of the frequency value to the total sample size is used as the real-time misjudgment probability value. The time window width T is dynamically adjusted according to the intensity of the competition: in the regular training mode, it is set to T=10 shooting events, and in the formal competition mode, it is set to T=5 shooting events. Environmental parameters are synchronously recorded into the statistical database, including ambient temperature, relative humidity, and ambient light intensity values.

[0116] The misjudgment probability risk assessment employs a multi-level verification process: When a single calculation shows a misjudgment probability ≥35% for a critical area, continuous monitoring mode is activated. The system checks whether this probability value remains ≥35% in the subsequent K consecutive shooting events, where K=3. When the continuous over-threshold condition is met, the probability counter's accumulated value reaches the trigger threshold, generating a high-precision grid verification command for the entire area. The command transmission process uses a redundant verification protocol, synchronously sending it to the grid control center through primary and backup dual channels.

[0117] Once the full-area high-precision grid verification mode is activated, the system immediately terminates the current multi-scale grid analysis process. All analysis sub-regions uniformly apply the highest-precision grid configuration.

[0118] The grid cell size was changed from the gradient distribution of the conventional mode (5mm-15mm) to a fixed value of 3mm across the entire domain;

[0119] The grid density parameter has been upgraded from the conventional grading (D1-D3) to a unified highest level, Dmax.

[0120] The time sampling interval is compressed from the dynamic range (12ms-30ms) to a fixed 8ms across the entire domain;

[0121] The mesh generation module performs a global synchronous reconstruction: based on the latest spatiotemporal coupling feature data, it generates a high-density mesh structure covering all monitoring areas around the basketball hoop in one go. The number of mesh nodes is increased to 2.8 times that of the conventional mode, and the accuracy of node spatial coordinate calculation is improved to the order of 0.1mm. The reconstructed mesh boundary processing adopts a globally unified coding rule to eliminate the splicing tolerance between the original sub-regions.

[0122] During operation in full-domain high-precision mode, the system monitors the computing resource load status in real time. The mode exit mechanism is triggered when any of the following conditions are met:

[0123] Complete global verification of M consecutive shooting events (M=3);

[0124] The probability of misjudgment in the key area drops back to the safe threshold (≤25%) for N consecutive shooting events (N=2).

[0125] System resource usage exceeds the safety threshold (CPU > 95% or memory > 90%).

[0126] Upon exiting the mode, an orderly degradation operation is performed: the global grid accuracy is reduced in stages within a single shooting event cycle: the first stage increases the grid cell size from 3mm to 5mm; the second stage restores the independent grid configuration for sub-regions; and the third stage reactivates the multi-scale strategy. Key monitoring area markers are automatically reset after degradation is complete, and historical misjudgment probability statistics for the area are cleared to zero.

[0127] The exception handling module intervenes in the following scenarios:

[0128] If the global mode runs for more than the preset time limit (180 seconds), it will be forcibly terminated and an exception report will be generated.

[0129] When two consecutive shooting events fail to generate valid trajectory data, the probability statistics will be automatically paused.

[0130] Sudden changes in environmental parameters (temperature change ≥5℃ / min or light change ≥300lux / s) trigger statistical calibration.

[0131] Data storage employs a sharded archiving strategy: In global mode, the complete grid data for each shot is stored in a high-throughput storage array, with added metadata tags including timestamps, environmental parameters, and verification conclusions. Historical operation logs record the time point, triggering reason, and peak resource consumption of mode switching events. Log files are stored in date-partitioned format with a 30-day retention period, supporting retrieval of historical mode switching records by time range.

[0132] The system provides real-time feedback on the operational status in high-precision mode across the entire domain to the monitoring terminal. The system generates a visual operational status graph.

[0133] The grid density thermal layer displays the calculated load distribution in each region;

[0134] The trend curve of misjudgment probability is overlaid with the resource consumption curve for display.

[0135] Mode switching events are marked on the monitoring interface in the form of timeline markers;

[0136] All status data is updated synchronously with shooting events, refreshing the displayed data after each shooting action. The monitoring terminal supports replay functionality, allowing users to review the global grid distribution pattern and misjudgment probability curves during any shooting event. Replay data loading latency is no more than 500 milliseconds, ensuring analysts have real-time access to the system's operational status.

[0137] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0138] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for preventing false positives in basketball goal detection using infrared time-series verification, characterized in that, include: The infrared signal sequence of the shooting action is collected in real time by an infrared sensor array deployed around the basketball hoop; the shooting trajectory segment containing the shooting trajectory is segmented from the infrared signal sequence. Spatiotemporal coupling features are extracted from the shooting trajectory segment; based on the spatiotemporal coupling features, a multi-scale trajectory analysis grid is generated according to the morphological features of the basketball trajectory; wherein, the grid accuracy corresponding to the trajectory region with drastic changes in infrared signal is higher than that of the region with gentle changes; Interference characteristics are analyzed on the time series of multi-scale trajectory analysis grids to generate an infrared signal interference index. Based on the infrared signal interference index, a dynamic path verification algorithm is used to calculate the abnormal deviation of the basketball trajectory; based on the comparison between the abnormal deviation and a preset threshold, a misjudgment verification conclusion of the shooting action is generated. The generated infrared signal interference index includes: Extract the spatiotemporal variation parameters of each grid node in the multi-scale trajectory analysis grid; The local anti-interference fluctuation index is calculated based on the fluctuation amplitude and correlation of spatiotemporal variation parameters; A global interference baseline is generated based on the local anti-interference fluctuation index of all grid nodes; Connect the output of the global interference baseline to the anomaly deviation calculation process; The calculation of abnormal deviation of the basketball trajectory using the dynamic path verification algorithm includes: Construct a standard motion path model for basketball trajectory; Dynamically correct the standard motion path model based on the infrared signal interference index; The instantaneous deviation is calculated by the deviation between the real-time trajectory coordinates and the corrected path model; The instantaneous deviation of the entire shooting process is used to generate a comprehensive abnormal deviation.

2. The infrared time-series verification method for basketball goal detection as described in claim 1, characterized in that, The process of segmenting the shooting trajectory segment containing the shooting trajectory from the infrared signal sequence includes: Identify the locations of all trough points in the infrared signal sequence; divide the infrared signal sequence into initial trajectory subsequences based on adjacent trough points; Calculate the signal intensity range of each initial trajectory subsequence; perform cluster analysis based on the signal intensity range of all initial trajectory subsequences to separate the effective shooting trajectory sequence; The output of the effective shooting trajectory sequence is connected to the spatiotemporal coupling feature extraction process.

3. The infrared time-series verification method for basketball goal detection to prevent false positives as described in claim 2, characterized in that, The process of generating a multi-scale trajectory analysis mesh based on the morphological characteristics of the basketball trajectory includes: Based on the spatial structure of the basketball hoop, the infrared sensor monitoring area is divided into multiple sub-regions for analysis. For each sub-region of analysis, the infrared signal sampling frequency is used to determine the grid cell size and grid density of that sub-region. Based on the aforementioned spatiotemporal coupling characteristics, independent grids are generated for each analysis sub-region. Mesh stitching is achieved by matching the mesh boundary nodes of adjacent analysis sub-regions using unique codes; Connect the output of the spliced ​​mesh to the interference feature analysis process.

4. The infrared time-series verification method for basketball goal detection to prevent false positives as described in claim 3, characterized in that, Also includes: Establish a confidence label library for shooting motions based on historical shooting data; When the overall abnormal deviation exceeds the preset threshold, the shooting action confidence label library is called for secondary verification. The secondary verification results are fed back into the misjudgment verification conclusion generation process.

5. The infrared time-series verification method for basketball goal detection as described in claim 4, characterized in that, The update of the shooting motion confidence label library based on historical shooting data includes: Extract the misjudgment type features from the misjudgment verification conclusion; Adjust the weighting coefficient of the shooting motion confidence label based on the characteristics of the misjudgment type; The updated weighting coefficients are then input into the secondary verification process.

6. The infrared time-series verification method for basketball goal detection to prevent false positives as described in claim 5, characterized in that, It also includes a dynamic resource allocation mechanism: Real-time monitoring of the infrared signal processing load of each analysis sub-region; Based on the mapping relationship between signal processing load and grid computing time, grid computing resources are dynamically allocated to high-load sub-regions; The resource allocation results are fed back to the independent grid generation process.

7. The infrared time-series verification method for basketball goal detection as described in claim 6, characterized in that, The execution of the dynamic resource allocation mechanism includes: When the signal processing load of a certain analysis sub-region exceeds the warning threshold, it is marked as a critical monitoring area; Reduce the grid precision in non-critical monitoring areas to free up computing resources; The freed-up computing resources will be reallocated to the grid for high-precision processing of critical monitoring areas.

8. The infrared time-series verification method for basketball goal detection as described in claim 7, characterized in that, Also includes: Statistical analysis of the probability of misjudgments in consecutive shots in key monitoring areas; When the false positive probability continues to exceed the risk threshold, the full-area high-precision grid verification mode is triggered. Connect the output of the full-area high-precision grid to the interference feature analysis process.

Images