CN-122024481-A - Double-lane outside geomagnetic vehicle track continuous reconstruction method and system for multiple miss detection

CN122024481ACN 122024481 ACN122024481 ACN 122024481ACN-122024481-A

Abstract

The invention discloses a double-lane outside geomagnetic vehicle track continuous reconstruction method and system for multiple miss detection, which are suitable for scenes in which geomagnetic sensors are arranged on two outside lane lines on two sides of a unidirectional double lane and the middle of the geomagnetic sensors are not arranged. And taking an event-level geomagnetic detection result as input, establishing a time index buffer memory ordered according to the measurement time stamp in the data center, and processing the out-of-order event in a fixed hysteresis sliding window. The method comprises the steps of outputting stable track segments, carrying out online self-calibration on the omission rate of each sensor group and obtaining reliability, constructing an event transfer diagram allowing to span a plurality of sensor groups, accumulating missing measurement penalties according to the reliability of an intermediate sensor group, combining space-time and feature consistency to form connection cost, and solving the minimum total cost connection under one-to-one constraint to obtain a track path. And generating a virtual track point for the cross-missing segment connection and fixing the lagged output stable track segment. The invention does not need waveform stage processing or additional hardware, and can improve track continuity and on-line output stability.

Inventors

MAO GUOQIANG
HOU ZHAOYING
LIU CHANG
MAO JIANQIANG
REN XIAOJIANG
LI CHANGLE

Assignees

西安电子科技大学

Dates

Publication Date: 20260512
Application Date: 20260228

Claims (9)

1. A double-lane outside geomagnetic vehicle track continuous reconstruction method for multiple miss detection is characterized by comprising the following steps: acquiring vehicle detection event reporting data of a plurality of groups of geomagnetic sensor networks arranged on lane lines on the two sides of a unidirectional double lane, wherein the vehicle detection event reporting data at least comprises a measurement time stamp, a sensor group identifier, a lane identifier and event-level geomagnetic characteristics; Maintaining a window of a time index cache in a data center, wherein the window of the time index cache covers a preset cache time window; Performing insertion and ordered maintenance of vehicle detection events that arrive out of order and whose measurement timestamps fall within windows of the time index buffer, and performing discard or bypass processing of an older event whose measurement timestamp is earlier than the beginning of a window of the time index buffer; merging the vehicle detection events in the window of the time index buffer memory into a scanning sequence according to the measurement time stamp, and maintaining a fixed hysteresis sliding window covering a plurality of continuous scans; Based on the outputted stable track segments, deducing expected passing times of the vehicle in each sensor group and counting actual detection times of the corresponding sensor group; according to the expected passing times and the actual detection times, the missing detection rate estimation of each sensor group is updated online, and the missing detection rate estimation is updated only by using the statistics of the output stable segments outside the window boundary; Constructing an event transfer graph in the fixed hysteresis sliding window, wherein nodes of the event transfer graph represent vehicle detection events, and edges of the event transfer graph represent candidate connection relations among the vehicle detection events, wherein the candidate connection relations are established when longitudinal position increment and equivalent speed constraint are met, and the event transfer graph is allowed to cross one or more intermediate sensor groups to represent multiple missed detection; Calculating a missing test penalty for the candidate connection relation according to the missing test rate estimation, and forming connection cost by the missing test penalty, space-time consistency cost and event feature consistency cost, wherein the missing test penalty sums up multiplication coefficients according to the reliability of a spanned intermediate sensor group , The more reliable the middle sensor group is, the larger the missing detection penalty is, and the missing detection penalty enables the connecting edge of the missing event at the reliable sensor group to have higher cost, so that error reconnection is restrained, and the lower penalty is allowed at the unreliable sensor group to improve the reconnection of the missing detection section; solving the minimum total cost connection relation of the event transfer graph under one-to-one constraint, so that each vehicle detection event is connected with at most one preceding event and one subsequent event, and generating a vehicle track path set according to the minimum total cost connection relation; When the minimum total cost connection relation spans one or more sensor groups or spans a plurality of scanning intervals, generating virtual track points according to measurement time stamps and longitudinal positions of events at two ends of the connection relation and writing track point sequences of corresponding vehicle tracks to form a missing segment interpolation result; and outputting a stable track section earlier than the fixed hysteresis sliding window boundary by adopting a fixed hysteresis output mechanism to obtain a continuously reconstructed vehicle track set, wherein the vehicle track set comprises a time stamp sequence, a longitudinal position sequence and a longitudinal speed sequence.
2. The multiple miss oriented dual lane outside geomagnetic vehicle trajectory continuous reconstruction method of claim 1, wherein the vehicle detection event includes a sensor identification and a quality indication quantity, and the event level geomagnetic characteristic includes at least one or more of a peak intensity, an event energy, and an event duration.
3. The multi-miss-oriented double-lane outside geomagnetic vehicle track continuous reconstruction method of claim 1, wherein the window maintenance rule of the time index buffer includes: the length of the preset buffer time window is as follows The current processing time of the data center is ; Measurement time stamp when a new arriving vehicle detects an event Satisfy the following requirements When the vehicle detection event is inserted into the time index buffer memory and ordered according to the measurement time stamp; When measuring time stamp Less than And when the vehicle detection event is judged to be an old event, discarding or bypassing offline statistical processing is performed.
4. The multi-miss-oriented double-lane outside geomagnetic vehicle track continuous reconstruction method of claim 1, wherein merging the vehicle detection events within the window of the time index buffer into a scan sequence according to a measurement timestamp includes: setting a scan interval Dividing the time axis into continuous scanning intervals ; Will fall into the same scan interval Within the vehicle detection event constitutes the first Sub-scan measurement set And forming the fixed hysteresis sliding window with an index of successive scans.
5. The multi-miss-oriented double-lane outside geomagnetic vehicle track continuous reconstruction method of claim 1, wherein updating the miss rate estimate of each sensor group on line includes: updating the omission factor estimation by adopting Beta-Bernoulli conjugate; And obtaining the reliability of each sensor group according to the omission ratio estimation, wherein the reliability is a monotonic function of the omission ratio estimation.
6. The multi-miss-oriented double-lane outside geomagnetic vehicle track continuous reconstruction method of claim 1, wherein the establishment constraint of the candidate connection relationship includes: The candidate connection relation satisfies that the longitudinal position of the subsequent event is larger than that of the precursor event, and the equivalent speeds of the subsequent event and the precursor event Satisfy the following requirements 。
7. The continuous reconstruction method of a double-lane outside geomagnetic vehicle track for multiple missed detection according to claim 1, wherein the calculation process of the missing detection penalty comprises: intermediate sensor group set spanned by candidate connection relation by missing detection penalty And accumulating, wherein the calculation of the missing measurement penalty meets the condition that the greater the reliability of the intermediate sensor group is, the greater the penalty is.
8. The multi-miss-oriented double-lane outside geomagnetic vehicle track continuous reconstruction method of claim 1, wherein solving the minimum total cost connection relation of the event transfer graph under one-to-one constraint comprises: And converting the event transfer graph into a bipartite graph formed by a precursor event set and a subsequent event set, and solving the minimum cost matching under one-to-one constraint to obtain the minimum total cost connection relation, or adopting a minimum component flow algorithm to obtain the vehicle track path set.
9. A double-lane outside geomagnetic vehicle track continuous reconstruction system for multiple miss detection is characterized by comprising: The data access module is used for accessing the vehicle detection event reported by the geomagnetic sensor; the time index buffer module is used for carrying out window interpolation sequencing on the vehicle detection events which arrive out of order and carrying out discarding or bypass processing on the old events; The scanning construction and sliding window module is used for merging the vehicle detection event into a scanning sequence and maintaining a fixed hysteresis sliding window; the self-calibration module of the omission factor is used for updating the omission factor estimation of each sensor group on line based on the stable track section; the event transfer graph construction module is used for constructing candidate connection relations which allow the connection relation to cross a plurality of sensor groups and calculating connection cost; the global connection solving module is used for solving the minimum total cost connection relation under one-to-one constraint and generating a vehicle track path set; and the interpolation and output module is used for generating a virtual track point at the missing position and outputting a stable track segment by adopting a fixed hysteresis output mechanism.

Description

Double-lane outside geomagnetic vehicle track continuous reconstruction method and system for multiple miss detection Technical Field The invention belongs to the technical fields of Intelligent Traffic Systems (ITS), traffic perception of the Internet of things and multi-target tracking, and particularly relates to a double-lane outside geomagnetic vehicle track continuous reconstruction method and system for multiple miss detection. Background The geomagnetic sensor has the advantages of low cost, simplicity and convenience in installation and maintenance, light exposure, small weather influence and the like, and is widely used for lane-level vehicle detection, vehicle speed estimation and traffic flow monitoring. In the deployment of partial engineering, in order to reduce construction damage and maintenance cost, geomagnetic sensors are often only arranged along the outer lane lines at two sides of a road, but not arranged on the middle lane line. The sensors typically upload the vehicle detection event to a roadside base station or data center via a wireless network. In the existing vehicle track tracking schemes, part of the schemes realize target tracking based on the fusion of radar and a camera, for example, patent literature (patent literature) a vehicle track tracking method and system based on the fusion of radar and video (application number: 202210867558.4, publication number: CN 115327527A) proposes to realize track tracking by utilizing multi-source sensor timestamp synchronization and data association. However, the scheme generally depends on the installation conditions such as a portal frame, the equipment cost is high, and the detection effect is easily influenced by the environment such as rain and fog. Another type of scheme realizes vehicle track tracking and track association based on a multi-geomagnetic sensor, for example, patent literature "a method for judging a false track based on vehicle track tracking of a multi-geomagnetic sensor" (application number: CN202410215061.3, publication number: CN118197070 a) proposes association and tracking based on geomagnetic detection results. However, in a scene laid outside the double lanes, geomagnetic detection event data often presents input characteristics of "sparse, missing detection, disordered sequence" in engineering, wherein multiple missing detection and continuous missing detection are important bottlenecks affecting track continuity. In addition, packet loss, congestion and retransmission of the wireless link may cause the arrival sequence of events to be inconsistent with the measurement time stamp, forming out-of-order input. If the 'waiting for alignment' or simple discarding strategy is adopted, larger delay is easily introduced or track information is lost, and real-time performance and global consistency are difficult to be simultaneously considered. Therefore, a new method and system that maintains track continuity under multiple miss conditions and reduces mis-association and ID switching while maintaining engineering landability is needed. Disclosure of Invention In order to solve the problems in the prior art, the invention provides a double-lane outside geomagnetic vehicle track continuous reconstruction method and system for multiple miss detection. The technical problems to be solved by the invention are realized by the following technical scheme: in a first aspect, the present invention provides a method for continuously reconstructing a geomagnetic vehicle track outside a double lane for multiple miss detection, including: Acquiring vehicle detection event reporting data of a plurality of groups of geomagnetic sensor networks arranged on lane lines on the two sides of a unidirectional double lane, wherein the vehicle detection event reporting data at least comprises a measurement time stamp, a sensor group identifier, a lane identifier and event-level geomagnetic characteristics; maintaining a window of a time index cache in a data center, wherein the window of the time index cache covers a preset cache time window; Performing insertion and ordered maintenance of vehicle detection events that arrive out of order and whose measurement timestamps fall within windows of the time index buffer, and performing discard or bypass processing of an older event whose measurement timestamp is earlier than the beginning of a window of the time index buffer; Merging the vehicle detection events in the window of the time index buffer memory into a scanning sequence according to the measurement time stamp, and maintaining a fixed hysteresis sliding window covering a plurality of continuous scans; estimating the miss rate of each sensor group on line according to the expected passing times and the actual detection times, and only using the stable segments which are output outside the window boundary for statistical updating; Constructing an event transfer graph in a fixed hysteresis sliding window, wherein nodes of the event trans