Research On Visual Detection System Of Vehicle Braking Time Sequence Based On Slip Rate Identification

In recent years,with the steady growth of China's economy and the rapid development of communications and transportation industry,road network and transportation facilities have been improved unprecedentedly,which has promoted the rapid development of the automotive industry and the continuous increase of Car Parc.The traffic safety issues that follow have become the focus of the whole society.Regular safety detection on various indicators of in-use vehicles is the main way to ensure vehicle safety,which the braking performance is the most important of all indicators.Especially for the semi-trailer train,as the main part of highway freight transportation,is developing towards multi-axles and heavy-duty.In view of its long body and complex structure,the braking performance indicators are of great significance.At present,there are two main methods to detect the braking performance of vehicles: road test and bench test.Road test must have a specific site,which is greatly affected by climatic conditions and has poor repeatability,so it is generally used as an auxiliary detection method.Bench test method occupies a small area,is scarcely affected by climatic conditions,and has good repeatability,which is a common method for braking detcetion in vehicle inspection stations and scientific institutions.The bench test method mainly relies on the anti-force roller brake tester and the plate brake tester,which can detect the braking force,braking unbalance and blocking force,so its can meets the detection of most types of vehicles.However,due to the large number of axles in tractor-semitrailers,the braking time sequence of each axle will have a direct impact on the braking stability of trains.If the front axle brakes fast,the vehicle can easily jack-knife at the moment of braking.Similarly,if the rear axle brakes fast,the vehicle can easily drag at the moment of braking.However,due to the limitation of testers' structure,the bench test method can not detect the braking time sequence of semi-trailer trains,so it is difficult to reflect the braking performance of the whole vehicle.Although the national standard GB 18565-2016 stipulates the testing method for braking time sequence of vehicle train,there is no relevant detection equipment at present due to the manufacturing cost and structure constraints,so the testing method has no practical significance.Therefore,it is an urgent need to develop a high precision and intelligent detection system for vehicle braking time sequence.With the deepening of the strategic deployment of Made-in-China 2025 and driven by the trend of industrial 4.0 with machine vision as the core,the field of vehicle detection is also moving towards the direction of informatization,automation and intellectualization.Therefore,based on the national standards and existing technical means,this paper introduces machine vision technology into vehicle braking time sequence detection,which proposes a detection method of vehicle braking time sequence based on stereo vision,designs and develops a visual detection system for vehicle braking time sequence.Based on the analysis of the instability mechanism for semi-trailer train and the influence of braking time sequence on braking stability,the causality of different braking time sequence is clarified.By analyzing the existing detection technology of braking time sequence,the direction and technical route of this paper are determined.The main research contents include the following four aspects:(1)Design of visual detection system for vehicle braking time sequenceAnalyzing the relationship between wheel slip rate and adhesion coefficient,the measurement target of braking time sequence is determined: the trigger time of brake pedal switch is taken as the initial timing scale,the time when the slip rate of each wheel reaches 20% is used as the detection result of braking time sequence.In order to identify wheel slip ratio,wheel slip ratio measurement model and correction model for curve driving based on visual measurement are established with white circular markers as indirect measurement object.Based on the principle of parallel binocular stereo vision measurement,deducing the structure model of the system,and the key factors affecting the comprehensive measurement error of the system are studied.Based on the above analysis,the visual detection system for vehicle braking time sequence is designed from three aspects: the overall layout,the detection process and the control scheme.(2)Research on key algorithms of image processingTo obtain the central coordinates of circular markers in image,the applicability of image processing algorithms is improved and optimized according to the characteristics of circular markers.Firstly,the original image is preprocessed,including image contrast enhancement,image deblurring,image denoising and image sharpening.Secondly,edge extraction is performed on normalized left and right images.In order to enhance the edge extraction effect of Canny algorithm for circular markers,the traditional Canny algorithm is improved in gradient direction expansion and adaptive threshold.To accurately identify circular markers,an ellipse detection method suitable for this study is proposed based on the existing ellipse detection theory.A boundary clearing algorithm is designed to remove redundant edges,and the accurate recognition and extraction of circular markers are realized by integrating least square theory and Hough transformation.Considering the large amount of computation in stereo matching for sequential images,a fast matching algorithm based on the epipolar constraint and Normalized Cross Correlation(NCC)algorithms is proposed.Finally,according to the three-dimensional(3D)reconstruction model and camera calibration parameters,3D reconstruction of the center coordinates of circular markers is carried out.(3)Research on calibration and accuracy verification of visual detection systemAnalyzing the linear and non-linear imaging models of cameras according to the transformation relationship among camera coordinate systems,the imaging model of camera calibration in this paper is determined.The principle and shortcomings of Zhang Zhengyou's calibration method are analyzed,an improved Zhang's camera calibration method based on PSO-LM(Particle Swarm Optimization and Levenberg-Marquardt)combined optimization strategy is proposed to realize nonlinear global optimization of calibration parameters,and the validity of the proposed method is verified by comparison of calibration experiments.In order to verify the dynamic recognition accuracy of circular markers by visual detection system,an accuracy verification device and method for simulating wheel braking are designed.Experiments of multi-cndition are carried out under multiple target speeds,and experimental errors in each case are analyzed.(4)Real vehicle test research on visual detection system of vehicle braking time sequenceIn order to verify the rationality of visual detection system design and the effectiveness of image processing algorithm and calibration algorithm,repeatability tests are carried out on the same semi-trailer train and universality tests for nine semi-trailer trains.In order to compare and analyze the measurement errors of the visual detection system,a set of wheel speed verifying device is designed by using wheel speed sensor on wheels and non-contact speed measuring instrument.The indication error and repeatability error of the two groups of test data are compared and analyzed,and the accuracy,stability and applicability of the detection system are verified.At the same time,in view of the uncertainty of measurement result error in the repetitive tests,the standard uncertainty is evaluated for the scientific evaluation of the detection system.Finally,the experimental error of visual detection system compared with calibration device is analyzed and summarized.