CN-122024498-A - Traffic flow detecting system based on pressure sensing

CN122024498ACN 122024498 ACN122024498 ACN 122024498ACN-122024498-A

Abstract

The invention relates to the technical field of traffic flow detection, and particularly discloses a traffic flow detection system based on pressure sensing. According to the invention, through collecting original signals of the multi-lane pressure sensor, performing filtering processing based on environmental state compensation, extracting pressure space profile distribution characteristics and pressure time sequence change rules, performing fusion analysis on the distribution characteristics and time sequence characteristics under the association of multi-sensor space layout, generating vehicle operation time-space characterization, inputting an adaptive classification model for vehicle type discrimination and flow statistics, solving the problem that the pressure signals are easy to be interfered by temperature and humidity and road vibration to cause false leakage detection, and solving the limitation that a fixed threshold model is difficult to adapt to multi-lane multi-vehicle type mixed running scenes and can not distinguish parallel vehicles from approaching conditions, thereby realizing the improvement of traffic flow detection precision and system self-adaption capability.

Inventors

CAO HAIGANG
WANG JIE
HAN BING
YAN ZHIYUAN
Xin Junchao
Tong Xingzhi
BAI YANG
ZHANG ZHONGLAN
ZHANG WENHAO
CHANG XIAOBING
YAN XIAOTING

Assignees

索睿邦(海南)科技发展有限公司

Dates

Publication Date: 20260512
Application Date: 20260210

Claims (10)

1. A traffic flow detection system based on pressure sensing, the system comprising: the pressure data acquisition module is used for acquiring original pressure signals output by a plurality of pressure sensors arranged in different lane detection areas of the road surface in real time; The signal anti-interference processing module is used for carrying out filtering processing based on environmental state compensation on the original pressure signal to generate an anti-interference stable pressure signal; the multi-mode feature extraction module is used for extracting distribution features representing the pressure space profile of the vehicle and time sequence features representing the time variation law of pressure from the stable pressure signal; The space-time fusion analysis module is used for carrying out fusion and association analysis on the distribution characteristics and the time sequence characteristics according to the spatial layout position relation of the pressure sensors, and generating vehicle running space-time characterization comprising complete pressure characteristic information of vehicles in time and space dimensions; the self-adaptive classification counting module is used for inputting the vehicle running time-space representation into a pre-trained self-adaptive classification model to judge the vehicle type, outputting a vehicle type judging result, and generating a flow statistical result corresponding to the lane and the vehicle type according to the vehicle type judging result.
2. The traffic flow detection system according to claim 1, wherein the filtering based on the environmental condition compensation performed by the signal anti-interference processing module is implemented by an adaptive nonlinear aggregation filtering algorithm that jointly optimizes signal quality in a time-frequency domain, the adaptive nonlinear aggregation filtering output signal Expressed as: , wherein, A stable pressure signal after interference resistance at the time t is represented, As a causal smoothing kernel function, Is a preset saturated nonlinear function which is a saturated nonlinear function, Representing the total number of pressure sensors, Representing the raw pressure signal value of the ith pressure sensor at time tau, Representing the time-varying trust weight of the ith pressure sensor at time tau, A noise estimation term representing the ith pressure sensor at time τ; the time-varying trust weight From the trend of ambient temperature And energy ratio of road vibration in critical frequency band The common regulation and control and the calculation mode are as follows: , wherein, And Is a parameter of sensitivity to be preset and is a parameter of sensitivity to be preset, , Is the time-frequency power spectrum of road surface vibration, And Is the upper and lower limit of the critical band.
3. The pressure sensing-based traffic flow detection system of claim 2, wherein the multi-modal feature extraction module is specifically configured to: the distribution features extracted from the steady pressure signal include a second moment matrix and a pressure gradient direction histogram of a pressure action region, the timing features including an empirical mode decomposition component of the pressure signal and its hilbert spectrum.
4. The pressure-sensing-based traffic flow detection system of claim 3, wherein the spatiotemporal fusion analysis module is specifically configured to: and constructing a space-time diagram model according to the spatial layout position relation of the pressure sensors, wherein nodes in the diagram correspond to the pressure sensors, the side weights represent space-time correlation among the sensors, and the distribution characteristics and the time sequence characteristics are aggregated through a diagram attention network to generate the vehicle running space-time characterization.
5. The pressure sensing-based traffic flow detection system of claim 4, wherein the adaptive classification model is a multi-head graph attention network with a gating loop unit for modeling time dependence of the vehicle run-time characterization, the multi-head graph attention network for capturing spatial interactions between multiple lanes.
6. The pressure sensing-based traffic flow detection system of claim 1, wherein the system further comprises: And the online self-learning module is used for dynamically adjusting the decision threshold of the self-adaptive classification model according to the deviation of the flow statistical result and the external traffic data.
7. The pressure sensing-based traffic flow detection system of claim 1 wherein the plurality of pressure sensors are arranged in a honeycomb grid within each lane, the grid cell side length being less than half a typical vehicle tire ground contact length.
8. The pressure sensing-based traffic flow detection system of claim 2, wherein the signal anti-tamper processing module is further configured to: after the stable pressure signal is generated, decomposing the stable pressure signal into a plurality of eigenmode functions through a variation mode decomposition algorithm, and selecting the eigenmode function with the largest energy as a signal segment representing a vehicle event.
9. The pressure sensing-based traffic flow detection system of claim 5, wherein the adaptive classification counting module is specifically configured to: And when the flow statistical result is generated, the parallel vehicle event and the vehicle-following approaching vehicle event are separated into different counting queues by calculating the mahalanobis distance and the time delay of the vehicle running time-space representation.
10. The pressure sensing-based traffic flow detection system of claim 4, wherein the spatiotemporal fusion analysis module uses a spatiotemporal tensor fusion function when generating the vehicle operation spatiotemporal characterization, the spatiotemporal tensor fusion function having the expression: , wherein, A third-order tensor representing the vehicle operation space-time representation, A matrix formed for the stack of distributed features of all pressure sensors, Representing the timing feature vector of the ith pressure sensor, And As a trainable spatial projection vector, Representing the Sigmoid activation function, Representing the operation of the vector outer product, And The dimensional parameters of the distribution feature and the timing feature, respectively.

Description

Traffic flow detecting system based on pressure sensing Technical Field The invention relates to the technical field of traffic flow detection, in particular to a traffic flow detection system based on pressure sensing. Background Along with the rapid development of intelligent traffic systems, traffic flow detection is an important link of road monitoring and management, and the accuracy and reliability of the traffic flow detection directly affect traffic scheduling efficiency. At present, a traffic flow detection scheme based on a pressure sensing technology is applied to partial scenes, and pressure signals when vehicles pass through are collected through pressure sensors buried in the road surface, so that vehicle counting and type identification are realized. However, the prior art still faces significant challenges in practical deployments. On the one hand, the pressure sensor is easy to be interfered by the environmental temperature, the humidity change and the road surface vibration, so that the signal stability is insufficient, the false detection and the missing detection phenomenon frequently occur in the peak period or under the complex vehicle condition, and the detection precision is difficult to continuously meet the high standard requirement of intelligent traffic. On the other hand, the existing system is used for processing pressure data and is dependent on a fixed threshold judgment and simple statistical model, and is lack of self-adaptive analysis capability for complex scenes such as multiple lanes, multiple vehicle types and the like, parallel vehicles and too close to the vehicles cannot be effectively distinguished, the depth and the intelligent level of data analysis are limited, and popularization and application of the system in high-flow scenes such as highways, urban arterial roads and the like are restricted. Accordingly, there is a need to provide a solution to the above-mentioned problems. Disclosure of Invention In order to solve the technical problems, the invention provides a traffic flow detection system based on pressure sensing, which has the following technical scheme: the pressure data acquisition module is used for acquiring original pressure signals output by a plurality of pressure sensors arranged in different lane detection areas of the road surface in real time; The signal anti-interference processing module is used for carrying out filtering processing based on environmental state compensation on the original pressure signal to generate an anti-interference stable pressure signal; the multi-mode feature extraction module is used for extracting distribution features representing the pressure space profile of the vehicle and time sequence features representing the time variation law of pressure from the stable pressure signal; The space-time fusion analysis module is used for carrying out fusion and association analysis on the distribution characteristics and the time sequence characteristics according to the spatial layout position relation of the pressure sensors, and generating vehicle running space-time characterization comprising complete pressure characteristic information of vehicles in time and space dimensions; the self-adaptive classification counting module is used for inputting the vehicle running time-space representation into a pre-trained self-adaptive classification model to judge the vehicle type, outputting a vehicle type judging result, and generating a flow statistical result corresponding to the lane and the vehicle type according to the vehicle type judging result. Further, the filtering processing based on the environmental state compensation, which is executed by the signal anti-interference processing module, is realized by an adaptive nonlinear aggregation filtering algorithm, the algorithm jointly optimizes the signal quality in a time-frequency domain, and the output signal of the adaptive nonlinear aggregation filtering is obtainedExpressed as: Wherein, the A stable pressure signal after interference resistance at the time t is represented,As a causal smoothing kernel function,Is a preset saturated nonlinear function which is a saturated nonlinear function,Representing the total number of pressure sensors,Representing the raw pressure signal value of the ith pressure sensor at time tau,Representing the time-varying trust weight of the ith pressure sensor at time tau,A noise estimation term representing the ith pressure sensor at time τ; the time-varying trust weight From the trend of ambient temperatureAnd energy ratio of road vibration in critical frequency bandThe common regulation and control and the calculation mode are as follows: Wherein, the AndIs a parameter of sensitivity to be preset and is a parameter of sensitivity to be preset,,Is the time-frequency power spectrum of road surface vibration,AndIs the upper and lower limit of the critical band. Further, the multi-modal feature extraction module is specifically configure