High-altitude parabolic positioning traceability analysis method and system

Abstract

The invention discloses a high-altitude parabolic positioning traceability analysis method and system, and particularly relates to the technical field of positioning traceability. Motion parameters of an object are calculated by adopting a multi-frame image analysis method, rotary motion is detected by utilizing a deep learning algorithm, if rotation exists, a Magnus effect compensation model is introduced to correct trajectory calculation, meanwhile, wind speed and wind direction data are collected in real time, airflow change is simulated in combination with a computational fluid dynamics model, and the influence of wind power on the trajectory is corrected. Based on the corrected trajectory data, the initial throwing position of the object is reversely deduced in combination with a building three-dimensional structure model, a target person in charge is determined, and a complete traceability report is generated by comparing an analysis result with a monitoring video, so that the accuracy and traceability of traceability analysis are ensured. The method can effectively reduce the calculation error caused by the rotation effect, wind field interference and other factors, improves the precision of high-altitude parabolic traceability, and provides a scientific basis for responsibility determination.

Description

Technical Field

[0001] The present invention relates to the technical field of positioning and traceability, and particularly to a method and system for analyzing the positioning and traceability of high-altitude parabolic objects. Background Art

[0002] Analyzing the positioning and traceability of high-altitude parabolic objects refers to determining the source of high-altitude parabolic objects and tracking the responsible persons through technical means and data analysis methods. This process usually combines video surveillance, physical measurement, intelligent algorithms, and big data analysis to improve the accuracy of traceability, ensure that relevant personnel bear corresponding responsibilities, and thus effectively prevent and reduce high-altitude parabolic behaviors and ensure public safety.

[0003] In practical applications, this analysis method mainly relies on technologies such as high-definition cameras, artificial intelligence (AI) image recognition, sound source localization, and three-dimensional modeling for comprehensive research and judgment. For example, the parabolic trajectory is captured by a camera, and the starting position of the object is calculated by combining factors such as air flow and gravity to accurately locate the source of the parabola. At the same time, some systems also use Internet of Things sensors and big data analysis to give early warnings for high-risk areas, improve governance efficiency, and reduce potential safety hazards.

[0004] The existing technologies have the following deficiencies:

[0005] During the process of tracing high-altitude parabolic objects, the spin and roll of the object will cause the Magnus effect, resulting in lateral drift of the trajectory, thus deviating from the traditional parabola motion model. In particular, lightweight objects (such as plastic bottles and paper boxes) are more easily affected by the airflow differences caused by rotation in the air, making their movement paths difficult to predict, while heavy objects (such as bricks and glass bottles) are less affected. Since the existing trajectory inversion algorithms usually only consider gravity and air resistance and do not fully model the rotation effect, this may lead to serious deviations in the calculated starting point of the parabola, thereby affecting the accuracy of determining the responsibility for high-altitude parabolic objects. Summary of the Invention

[0006] The purpose of the present invention is to provide a method and system for analyzing the positioning and traceability of high-altitude parabolic objects to solve the deficiencies in the background art.

[0007] To achieve the above purpose, the present invention provides the following technical solution: A method for analyzing the positioning and traceability of high-altitude parabolic objects, comprising the following steps:

[0008] S1: Capture the motion trajectory of the parabolic object at a high frame rate through high-definition cameras deployed on the exterior wall or surrounding area of the building, and use multi-angle cameras to cooperate in shooting to form three-dimensional perspective data;

[0009] S2: Use artificial intelligence image recognition technology to process the video data, extract the motion trajectory of the parabolic object, and perform target tracking through the Faster R-CNN algorithm;

[0010] S3: Adopt the multi-frame image analysis method to calculate the motion parameters of the object, including position, velocity, and acceleration. Through the deep learning algorithm, analyze the morphological characteristics of the object to determine whether there is rotational motion. If rotational motion is detected, introduce the Magnus effect compensation model and correct the trajectory calculation according to the rotational direction and angular velocity of the object.

[0011] S4: Collect real-time wind speed and wind direction data, and combine with the computational fluid dynamics model to simulate local airflow changes. Through aerodynamic calculations, correct the lateral offset of the object trajectory to improve the calculation accuracy of the parabolic starting point.

[0012] S5: Based on the corrected trajectory data, combine with the three-dimensional building structure model to back-calculate the initial throwing position of the object and determine the target responsible person.

[0013] S6: Compare the analysis results with the surveillance video to generate a complete traceability report, including trajectory reconstruction data, parabolic starting point calculation results, and target responsible person information.

[0014] Preferably, in S3, the Magnus effect compensation and trajectory correction include: calculating the Magnus force anomaly index according to the detected rotational angular velocity ω and rotational direction, specifically:

[0015] For each thrown object, after detecting its rotational angular velocity, construct a multi-dimensional feature vector: X = [ω, v, ΔX]; where: ω is the rotational angular velocity of the object, the rotational motion speed detected by image analysis, v is the linear velocity of the object, calculated from the position change of consecutive frames; ΔX represents the lateral offset of the object caused by the Magnus effect, and the deviation between the actual trajectory and the standard free-fall trajectory is calculated through trajectory fitting.

[0016] From the normal parabolic trajectory sample data, calculate the mean vector μ and covariance matrix Σ of the feature vector, and calculate the mean vector: where N is the number of normal samples, and X i is the feature vector of the i-th sample; calculate the covariance matrix: T is the matrix transpose;

[0017] For the parabolic event to be detected, the Mahalanobis distance between its feature vector X and the mean vector μ of the normal distribution is calculated as: where: (X - μ) is the deviation between the feature vector of the target thrown object and the normal mean vector, and Σ -1 represents the inverse matrix of the covariance matrix calculated from historical normal data;

[0018] According to the calculated Mahalanobis distance MD, construct the Magnus force anomaly index MFAI, which is used to measure the deviation degree of the current parabolic trajectory from the normal trajectory. The calculation formula of MFAI is as follows: Where: MD normal represents the average Mahalanobis distance of the statistically obtained normal samples, and MD max represents the set maximum abnormal Mahalanobis distance threshold.

[0019] Preferably, when 0 < MFAI < 1, it indicates that the rotation influence of the object is within the normal range, and the standard trajectory calculation method is maintained without additional compensation; when 1 ≤ MFAI ≤ 1, it indicates that the deviation caused by the Magnus effect is extremely serious and Magnus effect compensation correction is required.

[0020] Preferably, in combination with the object mass m, calculate the trajectory offset drift index caused by the Magnus effect and correct the trajectory. The method for obtaining the trajectory offset drift index is as follows:

[0021] Train the LSTM model, input the historical trajectory data X t = 9x t , y t , z t , v t , ω t ); x t , y t , z t represent the three-dimensional spatial position of the object at time t, v t represents the instantaneous velocity of the object at time t, and ω t represents the rotational angular velocity of the object at time t. Let the LSTM predict the next moment trajectory X t+1 , and calculate the error e t between the real trajectory and the predicted trajectory. The expression is: e t = ||X pred - X actual ||; X pred is the predicted trajectory, and X actual is the real trajectory. Calculate the trajectory offset drift index TDI, and the expression is: Where N is the number of time steps.

[0022] Preferably, if the trajectory offset drift index is greater than the set threshold, it is determined that the trajectory is greatly affected by the Magnus effect and there is a serious offset, and trajectory correction is required; otherwise, the trajectory is considered normal and no correction is required.

[0023] Preferably, in S3, deploy an ultrasonic anemometer or a lidar wind field detector in the monitoring area to collect wind speed and wind direction data in real time, and record the wind speed components of the wind speed in three-dimensional space;

[0024] Collect the 3D structural data of the buildings in the monitoring area, establish the computational domain, and use the finite element mesh generation technology to construct an unstructured mesh;

[0025] Set the inlet boundary conditions and input the measured wind speed and wind direction;

[0026] Set the outlet boundary conditions: set it as a free flow boundary to allow the wind to diffuse naturally in the computational domain;

[0027] Set the wall boundary conditions: adopt the no-slip boundary conditions for the building surface;

[0028] Solve the Navier-Stokes equations to simulate the air flow, calculate the wind field flow, and solve the equations: where: ρ is the air density, V is the wind speed field, p is the air flow pressure distribution, μ is the air viscosity coefficient, and F external is the external force;

[0029] Calculate the distribution of the wind speed in the entire computational domain, output the local air flow influence area, generate the wind speed streamline diagram, and judge whether the high-altitude parabolic trajectory is affected.

[0030] Preferably, calculate the influence of the wind force on the object, including:

[0031] The acceleration a of the object affected by the wind force wind , and the expression is: where, C d is the air resistance coefficient, and A is the windward area of the object;

[0032] Calculate the offset of the object trajectory, and the trajectory offset correction equation: where, T is the falling time of the object, and calculate the corrected trajectory: X ′ =X + ΔX, Y ′ =Y + ΔY;

[0033] Simulate the object trajectory and optimize the calculation of the parabolic starting point, including: numerically integrating the trajectory using the Runge-Kutta method, calculating the trajectory changes under different wind field conditions, and inversely calculating the original parabolic starting point in combination with the CFD simulation data.

[0034] Preferably, in S6, judge the trajectory matching degree by calculating the Euclidean distance error between the trajectory data and the monitored trajectory data to ensure that the trajectory calculation result is consistent with the monitoring video;

[0035] Verify the accuracy of the throwing position by calculating the parabolic starting point error and comparing the calculated parabolic starting point with the parabolic starting point identified in the monitoring video;

[0036] By comparing the target responsible person with the person detected by the surveillance video through the matching degree of the responsible person, the information of the responsible person is finally confirmed.

[0037] The present invention also provides a system for analyzing the positioning and tracing of high-altitude parabolic objects, including a monitoring and acquisition module, a target detection module, a trajectory analysis and rotation effect compensation module, an environmental impact correction module, a parabolic starting point reverse inference module, and a tracing report generation module;

[0038] Monitoring and acquisition module: Through high-definition cameras deployed on the exterior wall or surrounding area of the building, the movement trajectory of the parabolic object is captured at a high frame rate, and multi-angle cameras are used for collaborative shooting to form three-dimensional perspective data;

[0039] Target detection module: Using artificial intelligence image recognition technology, the video data is processed to extract the movement trajectory of the parabolic object, and target tracking is performed through the Faster R-CNN algorithm;

[0040] Trajectory analysis and rotation effect compensation module: Adopting a multi-frame image analysis method, the movement parameters of the object are calculated, including position, velocity, and acceleration. Through deep learning algorithms, the morphological characteristics of the object are analyzed to determine whether there is rotational motion; if rotational motion is detected, the Magnus effect compensation model is introduced to correct the trajectory calculation according to the rotation direction and angular velocity of the object;

[0041] Environmental impact correction module: Collect real-time wind speed and wind direction data, and combine with the computational fluid dynamics model to simulate local airflow changes. Through aerodynamic calculations, the lateral offset of the object trajectory is corrected to improve the calculation accuracy of the parabolic starting point;

[0042] Parabolic starting point reverse inference module: Based on the corrected trajectory data, combined with the three-dimensional building structure model, the initial throwing position of the object is reversed and the target responsible person is determined;

[0043] Tracing report generation module: Compare the analysis results with the surveillance video to generate a complete tracing report, including trajectory reconstruction data, parabolic starting point calculation results, and target responsible person information.

[0044] In the above technical solution, the technical effects and advantages provided by the present invention:

[0045] 1. The present invention acquires the motion trajectory of a thrown object with a high frame rate through a high-definition camera, and combines Faster R-CNN object detection and multi-camera data fusion to ensure the stability of target tracking. It uses multi-frame image analysis to calculate the motion parameters of the object, and detects rotational motion through a deep learning algorithm. If a rotational effect is detected, the Magnus force anomaly index (MFAI) is introduced for compensation and correction. Further, it combines LSTM time series prediction to calculate the trajectory deviation drift index (TDI) to determine whether the trajectory needs to be corrected, thereby enhancing the accuracy of tracing the trajectory of high-altitude parabolic objects.

[0046] 2. The present invention also combines real-time wind speed and direction data and a computational fluid dynamics (CFD) model to simulate local airflow changes and correct the influence of wind on the trajectory, optimizing the calculation of the parabolic starting point. In addition, by comparing the trajectory data with the surveillance video, the Euclidean distance error is used to calculate the degree of trajectory matching, ensuring that the calculation results are consistent with the actual situation, and the suspect is finally locked through the responsible person matching algorithm. Compared with the prior art, the present invention provides a more accurate method for reconstructing the parabolic trajectory, significantly improving the reliability and traceability of traceability analysis, providing a scientific basis for the liability determination of high-altitude parabolic cases, and helping to improve the level of public safety management. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings.

[0048] Figure 1 It is a flowchart of the method of the present invention.

[0049] Figure 2 It is a system module diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0050] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments of the present invention belong to the scope of protection of the present invention.

[0051] Example 1, please refer to Figure 1 As shown, a method for positioning and tracing the source of high-altitude parabolic objects in this embodiment includes the following steps:

[0052] S1: High-definition cameras deployed on the exterior walls or surrounding areas of buildings capture the motion trajectory of the thrown object at a high frame rate, and multi-angle cameras are used for collaborative shooting to form three-dimensional perspective data;

[0053] S2: Use artificial intelligence image recognition technology to process the video data, extract the motion trajectory of the thrown object, and perform object tracking through the Faster R-CNN algorithm;

[0054] S3: Adopt the multi-frame image analysis method to calculate the motion parameters of the object, including position, velocity, and acceleration. Through deep learning algorithms, analyze the morphological characteristics of the object to determine whether there is rotational motion; if rotational motion is detected, introduce the Magnus effect compensation model to correct the trajectory calculation according to the rotation direction and angular velocity of the object;

[0055] S4: Collect real-time wind speed and wind direction data, and combine with the computational fluid dynamics model to simulate local airflow changes. Through aerodynamic calculations, correct the lateral offset of the object trajectory to improve the calculation accuracy of the throwing starting point;

[0056] S5: Based on the corrected trajectory data, combined with the three-dimensional building structure model, reverse-deduce the initial throwing position of the object and determine the target responsible person;

[0057] S6: Compare the analysis results with the surveillance video to generate a complete traceability report, including trajectory reconstruction data, throwing starting point calculation results, and target responsible person information.

[0058] In S1, in the analysis of high-altitude throwing positioning and traceability, in order to accurately obtain the motion trajectory of the thrown object, the system needs to reasonably deploy high-definition cameras on the exterior walls or surrounding areas of buildings, and combine high frame rate and multi-angle camera technologies to achieve complete trajectory capture and three-dimensional perspective reconstruction.

[0059] First of all, the selection and parameter configuration of the camera are crucial. The deployed camera equipment should have a high-definition resolution (such as 4K or higher) to ensure that the shape, color, and motion details of the thrown object can be clearly captured. At the same time, in order to effectively detect high-speed moving objects, the camera should have a high frame rate (such as 60fps or 120fps) to reduce inter-frame blur and improve the accuracy of trajectory calculation. In addition, in night or low-light environments, infrared imaging or low-light enhancement technology can be combined to ensure all-weather monitoring capabilities.

[0060] Secondly, the installation position and angle design of the camera need to fully consider the occurrence probability of parabolic behavior and environmental factors. Usually, the camera should be installed on the exterior wall of the building, the top of the building, the square lamp post or public area, etc., to ensure that the possible parabolic areas are covered without dead angles. To achieve 3D trajectory reconstruction, multiple cameras need to cooperate to shoot from different angles. Among them, wide-angle cameras are used for large-scale monitoring, while PTZ (pan-tilt-zoom) cameras can automatically zoom in and track after detecting parabolic behavior to obtain clearer moving trajectory data. In addition, in areas with dense high-rise buildings, stereo vision technology can also be combined to calculate the depth information of objects using dual cameras to enhance the accuracy of trajectory positioning.

[0061] Finally, the fusion and processing of camera data are the key steps. The parabolic trajectory data captured by multi-angle cameras need to be synchronously processed to eliminate perspective errors, and computer vision technologies (such as inter-frame difference, object detection, and optical flow analysis) are used to extract the 3D motion trajectory of the thrown object. Combining these data, a complete parabolic path model can be constructed, which can then provide accurate input information for subsequent traceability calculations.

[0062] In S2, in the high-altitude parabolic positioning and traceability analysis, artificial intelligence image recognition technology is used to process video data, and the Faster R-CNN algorithm is combined for object tracking, including the following steps:

[0063] Step 1: Video data preprocessing: Collect video data captured by high-definition cameras to ensure that the resolution and frame rate meet the detection requirements (such as above 4K, 60fps). Perform video decoding and frame extraction to convert the video stream into a sequence of continuous image frames and perform time series synchronization processing for subsequent tracking calculations. Image enhancement technologies (such as denoising, contrast enhancement, gamma correction) are used to improve the visibility of the target object in complex environments and reduce background interference.

[0064] Step 2: Object detection and thrown object recognition: Use the Faster R-CNN object detection algorithm to perform object recognition on each frame of the image to detect possible thrown objects. The Region Proposal Network (RPN) is used to generate candidate target regions, and the Convolutional Neural Network (CNN) is used to extract features to classify thrown objects and non-thrown objects (such as birds, leaves). The Non-Maximum Suppression (NMS) algorithm is used to remove redundant detection boxes to ensure the accuracy and stability of object detection.

[0065] Step 3: Object Tracking and Trajectory Extraction: Use Faster R-CNN combined with Kalman Filter to track the thrown object, maintaining stable recognition of the object in consecutive frames. Calculate the pixel coordinates (x, y) of the object in the image and form a motion trajectory dataset by combining with time information (t). Further enhance the robustness of object tracking through the Optical Flow method or correlation matching algorithm to avoid tracking loss caused by short-term occlusion of the object.

[0066] Step 4: Trajectory Analysis and Parabolic Starting Point Calculation: Combine multi-frame object position data to fit the motion trajectory of the thrown object and determine whether it conforms to the free-fall motion pattern. Use methods such as second-order curve fitting or Bezier curve modeling to construct an accurate trajectory curve. Combine camera calibration data to convert pixel coordinates to actual physical space coordinates, and correct trajectory deviations by combining with the aerodynamic model, and finally calculate the position of the parabolic starting point.

[0067] Step 5: Anomaly Detection and Multi-Camera Data Fusion: Combine data from multiple cameras and use object matching algorithms (such as deep learning ReID technology) to ensure the correct association of the same object between different cameras. Calculate the trajectory consistency from different perspectives, eliminate false detection results, and optimize the motion trajectory of the thrown object. Use machine learning models (such as LSTM time series analysis) to detect abnormal motion patterns and identify possible false alarms or interfering objects.

[0068] In S3, multi-frame image analysis and motion parameter calculation include: Use a high-frame-rate camera (such as 60fps or 120fps) to capture the continuous motion trajectory of the thrown object and extract image frames at multiple time points. Perform object detection on the thrown object in each frame and calculate the center coordinates (x, y, t) of the object in the image through Faster R-CNN. Combine camera calibration data to convert pixel coordinates to actual physical space coordinates (X, Y, Z, t), and calculate the velocity v and acceleration a of the object. Use the second-order curve fitting method, such as the least squares method or Bezier curve, to estimate the initial motion trajectory of the object and obtain its main motion direction.

[0069] Rotational motion detection and Magnus effect judgment, including: analyzing the morphological changes of an object in multiple frames of images using deep learning algorithms (such as CNN+LSTM or ResNet), detecting its surface texture, edge features, and brightness changes, and judging whether rotational motion exists. If the object exhibits periodic morphological changes in consecutive frames (such as changes in stripe or mark positions), calculate its rotational angular velocity ω (unit: rad / s). Use the inter-frame difference method or Hough transform to analyze the deformation of the object's edge contour, further verifying the rotational direction and amplitude. Combine with the computational fluid dynamics (CFD) model to analyze the possible impact of rotational motion on the object's trajectory and judge whether Magnus effect compensation needs to be introduced.

[0070] Magnus effect compensation and trajectory correction, including: calculating the Magnus force anomaly index according to the detected rotational angular velocity ω and rotational direction, specifically:

[0071] For each projectile, after detecting its rotational angular velocity, construct a multi-dimensional feature vector to describe the parameters related to the Magnus effect: X = [ω, v, ΔX]; where: ω is the rotational angular velocity of the object (rad / s), the rotational motion speed detected by image analysis, v is the linear velocity of the object (m / s), calculated from the position changes in consecutive frames; ΔX represents the lateral offset (m) of the object caused by the Magnus effect, and the deviation between the actual trajectory and the standard free-fall trajectory is calculated by trajectory fitting.

[0072] From a large number of normal projectile trajectory sample data, calculate the mean vector μ and covariance matrix Σ of the feature vector. Calculate the mean vector: where, N is the number of normal samples, X i is the feature vector of the i-th sample; calculate the covariance matrix: T is the matrix transpose.

[0073] For the projectile event to be detected, the Mahalanobis distance between its feature vector X and the mean vector μ of the normal distribution is calculated as: where: (X - μ) is the deviation between the feature vector of the target projectile and the normal mean vector, Σ -1 represents the inverse matrix of the covariance matrix calculated from historical normal data.

[0074] According to the calculated Mahalanobis distance MD, construct the Magnus force anomaly index MFAI to measure the deviation degree of the current projectile trajectory from the normal trajectory. The calculation formula of MFAI is: where: MD normal represents the average Mahalanobis distance of the statistically obtained normal samples, MD maxRepresents the set maximum abnormal Mahalanobis distance threshold for normalization.

[0075] When 0 < MFAI < 1, it indicates that the rotational influence of the object is within the normal range, maintaining the standard trajectory calculation method without additional compensation; when 1 ≤ MFAI ≤ 1, it indicates that the deviation caused by the Magnus effect is extremely serious and Magnus effect compensation correction is required.

[0076] Combined with the object mass m, calculate the trajectory deviation drift index caused by the Magnus effect and correct the trajectory. The method for obtaining the trajectory deviation drift index is as follows:

[0077] Train the LSTM model, input historical trajectory data X t =(x t , y t , z t , v t , ω t ); x t , y t , z t represents the three-dimensional spatial position of the object at time t, v t represents the instantaneous velocity of the object at time t, ω t represents the rotational angular velocity of the object at time t. Let the LSTM predict the next moment trajectory X t+1 , and calculate the error e t between the real trajectory and the predicted trajectory. The expression is: e t =||X pred -X actual ||; X pred is the predicted trajectory, X actual is the real trajectory. Calculate the trajectory deviation drift index TDI, and the expression is: where N is the number of time steps.

[0078] If the trajectory deviation drift index is greater than the set threshold (such as 0.8), it is determined that the trajectory is greatly affected by the Magnus effect and there is a serious deviation, and trajectory correction is required. Otherwise, it is considered that the trajectory is normal and no correction is needed.

[0079] Adopt numerical integration methods (such as the Runge-Kutta method) to simulate the corrected object motion trajectory, and compare the correction results with the actual monitoring data to optimize the accuracy of trajectory calculation. Combine the high-rise building airflow data to adaptively adjust the influence range of the Magnus effect and improve the stability of trajectory back-calculation.

[0080] Generate a corrected three-dimensional trajectory coordinate point sequence and input it into the parabolic starting point calculation module. Combine multi-camera data fusion technology to ensure the coherence of trajectory calculation and improve the reliability of traceability analysis. Output the final parabolic starting point prediction result to provide an accurate basis for liability determination.

[0081] S4: Collect real-time wind speed and wind direction data, and combine with the computational fluid dynamics model to simulate local air flow changes. Through aerodynamic calculations, correct the lateral offset of the object's trajectory and improve the calculation accuracy of the parabolic starting point.

[0082] Deploy an ultrasonic anemometer or a lidar wind field detector in the monitoring area to collect wind speed Vw and wind direction θw data in real time. Combine with a meteorological database (such as the local meteorological bureau API) to obtain wind field information in a larger range to ensure data integrity. Set the wind speed data sampling frequency (such as 1Hz or higher) to ensure that wind field changes can be monitored in real time.

[0083] Record the wind speed vector V w =(V wx , V wy , V wz ), that is, the wind speed components in three-dimensional space. Record the wind direction θw and convert it into a wind speed vector for subsequent CFD simulation input. Use moving average filtering or Kalman filtering to process the wind speed data to eliminate sudden noise and improve data stability.

[0084] Collect the 3D structure data of the buildings in the monitoring area to establish a computational domain to ensure that the CFD simulation can accurately reflect the wind flow. Use finite element mesh generation technology to construct unstructured meshes (such as tetrahedral or hexahedral meshes) to ensure calculation accuracy.

[0085] Set the inlet boundary condition: Adopt the large eddy simulation (LES) or Reynolds-averaged Navier-Stokes (RANS) model and input the measured wind speed Vw and wind direction θw.

[0086] Set the outlet boundary condition: Set it as a free flow boundary to allow the wind to diffuse naturally in the computational domain.

[0087] Set the wall boundary condition: Adopt a no-slip condition on the building surface to ensure that the wind flow correctly bypasses the building.

[0088] Solve the Navier-Stokes equation to simulate the wind flow, calculate the wind field flow, and solve the equation: where: ρ is the air density (kg / m 3 ), usually taken as 1.225 kg / m 3, V is the wind speed field (m / s), p is the air flow pressure distribution (Pa), μ is the air viscosity coefficient (Pa·s), F external are external forces (such as Magnus effect and gravity).

[0089] Calculate the distribution of wind speed V(x,y,z) in the entire computational domain and output the local airflow impact area. Generate wind speed streamlines to identify areas where turbulence or updrafts may form and determine whether the trajectory of high-altitude parabolic objects is affected.

[0090] Calculate the effects of wind on objects, including:

[0091] The acceleration of the object affected by wind is a wind , the expression is: in, C d is the air resistance coefficient (set according to the shape of the object), A is the frontal area of ​​the object (m 2 ).

[0092] Calculate the offset of the object trajectory, trajectory offset correction equation: Where T is the falling time of the object. Calculate the corrected trajectory: X ′ =X+ΔX,Y ′ =Y+ΔY;

[0093] Simulate object trajectories and optimize parabola starting point calculations, including:

[0094] The Runge-Kutta method is used to numerically integrate the trajectory and calculate the trajectory changes under different wind field conditions. Combined with CFD simulation data, the original parabola starting point is reversely calculated to ensure that the trajectory calculation accuracy is improved.

[0095] Combine the camera target tracking data, wind speed and direction data, and CFD calculation results to comprehensively optimize the trajectory. Use particle filter (PF) or Kalman filter to dynamically correct the trajectory error. Set the wind speed impact threshold. For example, when Vw>5m / s, CFD correction data is preferred instead of standard parabola calculation. If the trajectory deviation before and after correction exceeds the normal range, it is marked as abnormal data, and the parabola starting point calculation method is further optimized.

[0096] S5: Based on the corrected trajectory data and combined with the three-dimensional structural model of the building, the initial throwing position of the object is inferred and the target responsible person is determined.

[0097] The corrected trajectory data includes the precise positions, velocities, and aerodynamic correction parameters of the object at different time points. To further determine the starting position of the parabola, the system matches the corrected trajectory with the three-dimensional structural model of the building to ensure the rationality of the trajectory in space. Through three-dimensional coordinate mapping, the motion state of the object at different time steps is calculated, and the reverse trajectory calculation method is used to trace back the throwing point of the object and align it with specific positions such as building floors, windows, or balconies.

[0098] During the reverse trajectory calculation process, the algorithm combines data from multi-angle cameras and uses multi-view geometry calculations to confirm whether the object was thrown from a specific window or balcony. For multiple possible starting points of the parabola, Bayesian probability inference or least squares optimization algorithms are used to calculate the most likely throwing position, and factors such as the parabola angle and hand movement pattern are combined to further narrow down the scope of tracing. In addition, considering the height of the building structure, window positions, and airflow effects, unreasonable throwing points are excluded to ensure that the calculated throwing positions conform to physical laws.

[0099] After locking the throwing position, the system further combines the face recognition and behavior analysis technologies of the intelligent monitoring system to trace back the personnel activities in the area. Through object detection algorithms, personnel near the window or balcony within a specific time are identified, and historical monitoring data is combined to determine whether there are suspicious behaviors, such as outward throwing actions and arm swing trajectories. In addition, household information and personnel entry and exit records can be combined to finally confirm the possible responsible person and generate an evidence report for use by property management, law enforcement agencies, or judicial organs, providing a scientific basis for tracing high-altitude throwing cases.

[0100] S6: Compare the analysis results with the surveillance video to generate a complete traceability report, including trajectory reconstruction data, parabola starting point calculation results, and target responsible person information.

[0101] Obtain the corrected object motion trajectory from the trajectory calculation model, including spatial coordinates, timestamps, and velocity information.

[0102] The trajectory data is represented as: T c ={(x i , y i , z i , t i )|i = 1, 2,..., P}; where: x i , y i , z i are the spatial coordinates of the object in the i-th frame, t i is the timestamp, and P is the total number of frames of the trajectory;

[0103] Extract the object motion trajectory data T in the surveillance video through object detection algorithms such as Faster R-CNN: v={(x′ i ,y′ i ,z′ i ,t′ i )|i = 1, 2, ..., Q}; where: x′ i ,y′ i ,z′ i is the position of the object detected in the monitoring screen, and t′ i is the timestamp in the video recording, and Q is the number of video frames.

[0104] The Euclidean distance error is used to calculate the deviation EDE between two trajectories:

[0105] If EDE < 0.2m, the trajectory matching degree is high, and it is determined that the calculated trajectory is consistent with the monitoring video; if EDE > 0.5m, it indicates that there is an error, and the trajectory calculation model needs to be adjusted or the monitoring video needs to be re-analyzed.

[0106] Compare the calculated result of the parabolic starting point with the monitoring video: The parabolic starting point calculated by backtracking the trajectory: P c =(x s ,y s ,z s ,t s ); where: x s ,y s ,z s are the coordinates of the parabolic starting point, and t s is the time when the object leaves the hand.

[0107] In the monitoring video, use time synchronization to match the detected throwing action: P v =(x′ s ,y′ s ,z′ s ,t′ s ); Calculate the deviation of the parabolic starting point position: If EP < 0.3m, the calculation result is credible, and the parabolic starting point matches the monitoring video; if EP > 0.5, the trajectory correction parameters or the time deviation of the monitoring video need to be checked.

[0108] Combine the floor of the parabolic starting point, monitoring data, and household information to screen out the target responsible person: R c ={ID1, ID2, ..., ID k}; where: ID k is the target responsible person speculated by the system. According to the behavior analysis algorithm, it is calculated whether there is any activity in the household area during the parabolic time.

[0109] Lock the person who actually threw the object in the monitoring through face recognition and behavior analysis: Rv = {ID'1, ID'2,..., ID' m}。

[0110] Calculate the matching degree between the target responsible person and the monitoring and identification responsible person: If P R > 0.8, it means that the calculated inference result is consistent with the monitoring, and the determination of the responsible person is successful. If P R <0.5, additional evidence (such as fingerprints, resident visitor records) needs to be combined for confirmation.

[0111] In this embodiment, first, high-definition cameras are deployed on the exterior wall or surrounding area of the building to capture the motion trajectory of the thrown object at a high frame rate, and multi-angle collaborative shooting is used to form three-dimensional perspective data. Then, using artificial intelligence image recognition technology, the video data is processed through the Faster R-CNN algorithm to extract and track the trajectory of the thrown object. Further, a multi-frame image analysis method is adopted to calculate the motion parameters of the object, and a deep learning algorithm is combined to detect rotational motion. If there is a rotational effect, the Magnus effect compensation model is introduced to correct the trajectory calculation. Next, real-time wind speed and wind direction data are collected, and the computational fluid dynamics (CFD) model is combined to simulate local airflow changes to correct the influence of wind force on the trajectory and improve the calculation accuracy of the throwing starting point. Subsequently, based on the corrected trajectory data, combined with the three-dimensional building structure model, the initial throwing position of the object is deduced backward, and the target responsible person is determined. Finally, the analysis results are compared with the surveillance video to generate a complete traceability report, including trajectory reconstruction data, throwing starting point calculation results, and responsible person information, providing scientific and effective evidence support for law enforcement agencies.

[0112] Embodiment 2, please refer to Figure 2 As shown, the high-altitude throwing positioning and traceability analysis system described in this embodiment includes a monitoring and acquisition module, a target detection module, a trajectory analysis and rotation effect compensation module, an environmental impact correction module, a throwing starting point backward deduction module, and a traceability report generation module;

[0113] Monitoring and acquisition module: Through high-definition cameras deployed on the exterior wall or surrounding area of the building, the motion trajectory of the thrown object is captured at a high frame rate, and multi-angle camera collaborative shooting is used to form three-dimensional perspective data;

[0114] Target detection module: Using artificial intelligence image recognition technology, the video data is processed to extract the motion trajectory of the thrown object, and target tracking is performed through the Faster R-CNN algorithm;

[0115] Trajectory Analysis and Rotation Effect Compensation Module: Adopt the multi-frame image analysis method to calculate the motion parameters of the object, including position, velocity, and acceleration. Through deep learning algorithms, analyze the morphological characteristics of the object to determine whether there is rotational motion. If rotational motion is detected, introduce the Magnus effect compensation model to correct the trajectory calculation according to the rotation direction and angular velocity of the object.

[0116] Environmental Impact Correction Module: Collect real-time wind speed and wind direction data, and combine with the computational fluid dynamics model to simulate local airflow changes. Through aerodynamic calculations, correct the lateral offset of the object trajectory to improve the calculation accuracy of the parabolic starting point.

[0117] Parabolic Starting Point Inverse Deduction Module: Based on the corrected trajectory data, combine with the three-dimensional building structure model to inverse deduce the initial throwing position of the object and determine the target responsible person.

[0118] Traceability Report Generation Module: Compare the analysis results with the surveillance video to generate a complete traceability report, including trajectory reconstruction data, parabolic starting point calculation results, and target responsible person information.

[0119] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data for software simulation to get a formula closest to the actual situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0120] It should be understood that the term "and / or" in this article is only a correlation relationship describing related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, and B exists alone. Here, A and B can be singular or plural. In addition, the character " / " in this article generally represents an "or" relationship between the related objects before and after, but it may also represent an "and / or" relationship. Specifically, it can be understood by referring to the context before and after.

[0121] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed in this article can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but this implementation should not be considered to exceed the scope of this application.

[0122] The above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed in this application can easily think of changes or substitutions, which should all be covered within the protection scope of this application.

Claims

1. A method for analyzing the positioning and tracing of high-altitude parabolic objects, characterized in that: It includes the following steps: S1: Use high-definition cameras deployed on the exterior walls or surrounding areas of buildings to capture the motion trajectory of the thrown object at a high frame rate, and use multi-angle cameras to cooperate in shooting to form three-dimensional perspective data; S2: Use artificial intelligence image recognition technology to process the video data, extract the motion trajectory of the thrown object, and perform target tracking through the Faster R-CNN algorithm; S3: Adopt a multi-frame image analysis method to calculate the motion parameters of the object, including position, velocity, and acceleration. Through deep learning algorithms, analyze the morphological characteristics of the object to determine whether there is rotational motion; If rotational motion is detected, introduce the Magnus effect compensation model to correct the trajectory calculation according to the rotation direction and angular velocity of the object; S4: Collect real-time wind speed and wind direction data, and combine with the computational fluid dynamics model to simulate local airflow changes. Through aerodynamic calculations, correct the lateral offset of the object trajectory to improve the calculation accuracy of the throwing starting point; S5: Based on the corrected trajectory data, combine with the three-dimensional building structure model to reverse the initial throwing position of the object and determine the target responsible person; S6: Compare the analysis results with the surveillance video to generate a complete traceability report, including trajectory reconstruction data, calculation results of the throwing starting point, and information on the target responsible person.

2. The method for analyzing the positioning and tracing of high-altitude parabolic objects according to claim 1, characterized in that: In S3, the Magnus effect compensation and trajectory correction include: calculating the Magnus force anomaly index according to the detected rotational angular velocity ω and rotation direction. Specifically: For each thrown object, after detecting its rotational angular velocity, construct a multi-dimensional feature vector: X = [ω, v, ΔX]; where: ω is the rotational angular velocity of the object, the rotational motion speed detected by image analysis, v is the linear velocity of the object, calculated from the position change of consecutive frames; ΔX represents the lateral offset of the object caused by the Magnus effect, and the deviation between the actual trajectory and the standard free-fall trajectory is calculated through trajectory fitting; From the sample data of the normal parabolic trajectory, calculate the mean vector μ and covariance matrix Σ of the feature vectors. Calculate the mean vector: where N is the number of normal samples, and X i is the feature vector of the i-th sample; calculate the covariance matrix: T is the matrix transpose; For the parabolic event to be detected, the Mahalanobis distance between its feature vector X and the mean vector μ of the normal distribution is calculated as follows: Where: (X - μ) is the deviation between the feature vector of the target parabolic object and the normal mean vector, and Σ -1 represents the inverse matrix of the covariance matrix calculated from historical normal data; According to the calculated Mahalanobis distance MD, construct the Magnus force anomaly index MFAI to measure the deviation degree of the current parabolic trajectory from the normal trajectory. The calculation formula of MFAI is as follows: Where: MD normal represents the average Mahalanobis distance of the statistically obtained normal samples, and MD max represents the set maximum abnormal Mahalanobis distance threshold.

3. The method for analyzing the positioning and tracing of high-altitude parabolic objects according to claim 2, characterized in that: When 0 < MFAI < 1, it means that the rotational influence of the object is within the normal range, maintain the standard trajectory calculation method, and no additional compensation is required; when 1 ≤ MFAI ≤ 1, it means that the offset caused by the Magnus effect is extremely serious, and Magnus effect compensation correction is required.

4. A method for analyzing the positioning and tracing of high-altitude parabolic objects according to claim 1, characterized in that: Combine the mass m of the object to calculate the trajectory offset drift index caused by the Magnus effect and correct the trajectory. The method for obtaining the trajectory offset drift index is: Train the LSTM model with the input of historical trajectory data X t = x t , y t , z t , v t , ω t ; x t , y t , z t represent the three-dimensional spatial position of the object at time t, v t represents the instantaneous velocity of the object at time t, ω t represents the rotational angular velocity of the object at time t. Let the LSTM predict the next moment trajectory X t+1 , and calculate the error e between the true trajectory and the predicted trajectory t , and the expression is: e t = X pred - X actual ; X pred is the predicted trajectory, X actual is the true trajectory. Calculate the trajectory deviation drift index TDI, and the expression is: where N is the number of time steps.

5. The method for analyzing the positioning and tracing of high-altitude parabolic objects according to claim 4, wherein: If the trajectory offset drift index is greater than the set threshold, it is judged that the trajectory is greatly affected by the Magnus effect and there is a serious offset, and trajectory correction is required; otherwise, it is considered that the trajectory is normal and no correction is required.

6. The method for analyzing the positioning and tracing of high-altitude parabolic objects according to claim 1, wherein: In S3, deploy ultrasonic anemometers or lidar wind field detectors in the monitoring area to collect real-time wind speed and wind direction data, and record the wind speed components of the wind speed in three-dimensional space; Collect 3D structure data of the buildings in the monitoring area, establish a computational domain, and use finite element mesh generation technology to construct an unstructured mesh; Set the inlet boundary conditions and input the measured wind speed and wind direction; Set the outlet boundary conditions: set as a free-flow boundary to allow the wind to naturally diffuse in the computational domain; Set the wall boundary conditions: Apply the no-slip boundary condition to the building surface; Solve the Navier-Stokes equations to simulate the air flow, calculate the wind field flow, and solve the equations: where: ρ is the air density, V is the wind speed field, p is the air pressure distribution, μ is the air viscosity coefficient, and F external is the external force; Calculate the distribution of wind speed in the entire computational domain, output the local airflow influence area, generate a wind speed streamline diagram, and determine whether the trajectory of high-altitude object throwing is affected.

7. The method for analyzing the positioning and tracing of high-altitude parabolic objects according to claim 6, characterized in that: Calculate the influence of wind force on the object, including: The acceleration a of the object affected by the wind force wind , and the expression is:[[]] Where,[[]] C d is the air resistance coefficient, and A is the windward area of the object; Calculate the offset of the object's trajectory. Trajectory offset correction equation: where T is the object's falling time, and calculate the corrected trajectory: X ′ = X + ΔX, Y ′ = Y + ΔY; Simulate the object trajectory and optimize the calculation of the throwing starting point, including: Numerically integrate the trajectory using the Runge-Kutta method, calculate the trajectory changes under different wind field conditions, and combine with CFD simulation data to perform inverse calculation on the original throwing starting point.

8. A method for positioning, tracing and analyzing high-altitude parabolic objects according to claim 7, characterized in that: In S6, judge the trajectory matching degree by calculating the Euclidean distance error between the trajectory data and the monitored trajectory data to ensure that the trajectory calculation result is consistent with the surveillance video; Verify the accuracy of the throwing position by calculating the throwing starting point error and comparing the calculated throwing starting point with the throwing starting point identified in the surveillance video; Finally confirm the responsible person information by comparing the target responsible person with the person detected in the surveillance video through the responsible person matching degree.

9. An aerial throwing positioning and traceability analysis system for implementing the aerial throwing positioning and traceability analysis method according to any one of claims 1-8, characterized in that: Including a monitoring and acquisition module, a target detection module, a trajectory analysis and rotation effect compensation module, an environmental impact correction module, a throwing starting point inverse deduction module, and a traceability report generation module; Monitoring and acquisition module: Through high-definition cameras deployed on the building exterior wall or surrounding area, capture the movement trajectory of the thrown object at a high frame rate, and use multi-angle cameras to cooperate in shooting to form three-dimensional perspective data; Target detection module: Use artificial intelligence image recognition technology to process the video data, extract the movement trajectory of the thrown object, and perform target tracking through the Faster R-CNN algorithm; Trajectory analysis and rotation effect compensation module: Adopt a multi-frame image analysis method to calculate the movement parameters of the object, including position, speed, and acceleration, analyze the morphological characteristics of the object through a deep learning algorithm, and judge whether there is rotational movement; If rotational movement is detected, introduce the Magnus effect compensation model to correct the trajectory calculation according to the rotation direction and angular velocity of the object; Environmental impact correction module: Collect real-time wind speed and wind direction data, and combine with the computational fluid dynamics model to simulate local airflow changes, and correct the lateral deviation of the object trajectory through aerodynamic calculations to improve the calculation accuracy of the throwing starting point; Throwing starting point inverse deduction module: Based on the corrected trajectory data, combine with the building three-dimensional structure model to inverse deduce the initial throwing position of the object and determine the target responsible person; Traceability report generation module: Compare the analysis results with the surveillance video to generate a complete traceability report, including trajectory reconstruction data, throwing starting point calculation results, and target responsible person information.

Images