Search

CN-121980506-A - Multi-source track data fusion processing method for real-time monitoring

CN121980506ACN 121980506 ACN121980506 ACN 121980506ACN-121980506-A

Abstract

The invention discloses a multi-source track data fusion processing method oriented to real-time monitoring, which relates to the technical field of data fusion processing and comprises the steps of collecting and uniformly preprocessing multi-source heterogeneous data in real time; the method comprises the steps of carrying out logical matching between time and space and a lane, accurately associating anonymous events detected in a tunnel to a specific vehicle, reversely deducing the most probable track segments from an entrance to the event points by taking the event points as strong constraint anchor points, dynamically distributing confidence coefficient weights along time variation to track data of various sources and reverse deduction results, generating a continuous global track penetrating through the front and the rear of the event based on time and space alignment and confidence coefficient weighted fusion, outputting the fused track and automatically triggering monitoring and early warning based on the event and track behaviors. According to the invention, by introducing a double-drive processing mechanism combining event-triggered reverse track deduction and multi-source forward fusion, the vehicle track can be reconstructed under the specific scene that signals such as expressway tunnel groups are limited and sudden events are easy to occur.

Inventors

  • CAO JINGBO
  • PAN YUE

Assignees

  • 湖州阳马科技有限公司

Dates

Publication Date
20260505
Application Date
20260122

Claims (9)

  1. 1. The multi-source track data fusion processing method for real-time monitoring is characterized by comprising the following steps of: receiving in real-time a target track data stream from at least two independent physical sensor systems and receiving an event report data stream from a fixed point event detection device; When the event report data stream indicates that a triggering event of a preset type occurs at a specific space position, carrying out association matching on the triggering event and a target individual related to the target track data stream to form an association event with an individual identifier; taking the occurrence position and time of the related event as terminal space-time constraint, combining with vehicle dynamics constraint and road topology structure, and calculating the most probable motion track segment from the entrance of the target individual to the occurrence time of the related event by a reverse deduction model, which is called a reverse deduction track; Respectively and dynamically distributing confidence values for different data points in the target track data stream and track points in the reverse deduction track according to the reliability of data sources, the space-time distance between the data points and the associated events and the environmental characteristics; Based on a time axis, carrying out weighted fusion calculation on track data points from different sources at the same moment and corresponding confidence values thereof to generate a continuous and fused global track; and outputting the global track and information related to the associated event.
  2. 2. The multi-source track data fusion processing method for real-time monitoring according to claim 1, wherein the at least two independent physical sensor systems comprise a first sensor subsystem which is deployed at an entrance of a monitoring area and is used for collecting an entrance state of a target, and a second sensor subsystem which can provide global position information of the target, and the fixed point event detection equipment is deployed inside the monitoring area.
  3. 3. The method for processing multi-source track data fusion oriented to real-time monitoring according to claim 1, wherein the performing association matching on the trigger event and the target individual specifically comprises: Maintaining a list containing the target individuals recently entering the monitoring area and their entering status information; Screening one or more candidate target individuals which enter before the occurrence time and the entry position of which accords with a preset space association rule with the occurrence position of the event from the list according to the occurrence time and the position of the trigger event; If a plurality of candidate target individuals exist, further screening is carried out according to the interval between the entry time of the candidate target individuals and the event occurrence time and the theoretical minimum transit time from the entry position to the event occurrence position so as to determine the uniquely associated target individuals.
  4. 4. The method for processing multi-source trajectory data fusion oriented to real-time monitoring according to claim 1, wherein the reverse deduction model is realized by constructing and optimizing an objective function, and the objective function fuses the following constraints: kinematic consistency constraint ensures that the position change at adjacent moments accords with the speed; The event terminal constrains, and the occurrence time and the position of the terminal point of the forced deduction track at the associated event meet the motion state corresponding to the event type; Track smoothness constraint, namely inhibiting abrupt change of track speed; the weight coefficients used for balancing the kinematic consistency constraint, the event terminal constraint and the track smoothness constraint in the objective function are determined through calibration of an offline optimization process based on historical precision data of the physical sensor system and a vehicle dynamics model.
  5. 5. The method for processing multi-source track data fusion oriented to real-time monitoring according to claim 1, wherein the method is characterized by dynamically distributing confidence values to track points in the reverse deduction track, and specifically comprises the following steps: assigning the highest-level initial confidence value to the track point at the occurrence time of the associated event; And carrying out attenuation calculation on the track points before the occurrence time of the related event according to the space distance between the track points and the occurrence position of the event and the line-shaped tortuosity degree of the road between the track points, wherein the confidence degree attenuation is larger when the track points are far away and the path is tortuosity.
  6. 6. The method for processing multi-source track data fusion oriented to real-time monitoring according to claim 1, wherein the weighted fusion calculation adopts a weighted average algorithm based on confidence values, and for a plurality of track data points existing in the same time stamp, the fusion result is a weighted average value of the position information of each data point by taking the confidence value as a weight.
  7. 7. The method for processing multi-source track data fusion oriented to real-time monitoring according to claim 1, wherein the step of generating the global track further comprises: and regarding the time period after the occurrence of the association event, taking the occurrence position and time of the association event as the initial anchor point of the subsequent track estimation, and adopting a motion state limiting model corresponding to the event type to carry out forward track estimation until new effective sensor data are acquired.
  8. 8. The method for processing multi-source track data fusion oriented to real-time monitoring according to claim 2, further comprising a data preprocessing step after the real-time receiving step and before the association matching step: Unifying the time stamps of all input data to the same time reference; and converting the space coordinates of all input data into a one-dimensional longitudinal distance coordinate system and a transverse offset coordinate system which take the entrance of the monitoring area as a reference origin.
  9. 9. The method for processing multi-source track data fusion oriented to real-time monitoring according to claim 1, wherein the outputting step further comprises automatically generating and sending abnormal state early warning information when the global track shows that the stay time of the target individual in a non-allowed area exceeds a preset threshold.

Description

Multi-source track data fusion processing method for real-time monitoring Technical Field The invention belongs to the technical field of data fusion processing, and particularly relates to a multi-source track data fusion processing method for real-time monitoring. Background In the field of real-time vehicle monitoring of expressway tunnel groups, multi-source track data fusion is a key technology for ensuring continuous and accurate positioning of vehicles. Existing methods generally rely on forward prediction and fusion mechanisms from history or current time to future, such as integration and trajectory smoothing of data sources from entry radar, intermittent satellite positioning, etc., by dynamic weighted averaging, kalman filtering, or their variant algorithms. The techniques are based on the assumption that the motion state (such as speed and acceleration) of the vehicle keeps continuous or regular change in a short time, and can achieve better effect in the scene of normal uniform speed or uniform acceleration running of the vehicle. However, in a specific and closed environment of the tunnel group, when a sudden emergency (such as a fault stop or an abnormal low-speed running) occurs, the movement mode of the vehicle can be instantaneously deviated from the continuous variation premise depending on the conventional forward prediction model. The event causes a non-patterned abrupt change in the trajectory, and the shielding of the satellite signals by the tunnel structure in turn causes a lack or unreliability of positioning data in the critical area. At this point, existing data fusion methods based primarily on forward extrapolation face the fundamental challenge that they lack the ability to reverse-infer and backtrack calibration for such incidents. The system can not only effectively utilize sparse fixed point event detection information near an event occurrence point, but also can not correct track prediction deviation generated by model hypothesis failure, so that the generated fusion track is interrupted at a key moment or is inconsistent with an actual path of a vehicle, and situation perception reliability of the monitoring system in an emergency scene is reduced. Therefore, the following means are proposed for solving the above problems. Disclosure of Invention The invention aims to provide a multi-source track data fusion processing method oriented to real-time monitoring, which can reconstruct a vehicle track continuously and reliably under a specific scene of limited signals such as a highway tunnel group and the like and easy occurrence of an emergency through introducing a double-drive processing mechanism combining event-triggered reverse track deduction and multi-source forward fusion, and solves the problem that the track reconstruction is easy to interrupt or misalign when the existing fusion method mainly relying on forward prediction faces discontinuous and irregular mutation of a vehicle motion mode caused by the emergency, thereby influencing the situation perception effectiveness of a monitoring system. In order to solve the technical problems, the invention is realized by the following technical scheme: the invention discloses a multi-source track data fusion processing method oriented to real-time monitoring, which comprises the following steps: receiving in real-time a target track data stream from at least two independent physical sensor systems and receiving an event report data stream from a fixed point event detection device; When the event report data stream indicates that a triggering event of a preset type occurs at a specific space position, carrying out association matching on the triggering event and a target individual related to the target track data stream to form an association event with an individual identifier; taking the occurrence position and time of the related event as terminal space-time constraint, combining with vehicle dynamics constraint and road topology structure, and calculating the most probable motion track segment from the entrance of the target individual to the occurrence time of the related event by a reverse deduction model, which is called a reverse deduction track; Respectively and dynamically distributing confidence values for different data points in the target track data stream and track points in the reverse deduction track according to the reliability of data sources, the space-time distance between the data points and the associated events and the environmental characteristics; Based on a time axis, carrying out weighted fusion calculation on track data points from different sources at the same moment and corresponding confidence values thereof to generate a continuous and fused global track; and outputting the global track and information related to the associated event. Further, the at least two independent physical sensor systems comprise a first sensor subsystem which is deployed at the entrance of the monitoring area and