| Road-rail vehicle is a special vehicle which relies on the coordinated operation of highways and railways.It has the characteristics of strong mobility of highway transportation and large carrying capacity of railway transportation.When the road-rail vehicle is running from road to railway,it is necessary to align the steel wheels with the rail before putting the Wheelsets on the track and realizing the conversion to railway.At present,in the freight yard,the horizontal distance between steel wheel and rail and the orientation information of vehicle are manually observed to guide the tractor driver to align the track.This method is not only inefficient,but also has poor guidance effect,which is difficult to meet the application and development needs of road-rail vehicle.Based on this,combined with machine vision technology,this paper intends to develop an intelligent wheelset drop guidance system,in order to replace the current manual assisted guidance mode.The system can real-time online detect the working conditions under the vehicle,accurately measure the horizontal distance between the steel wheel and rail and the deviation angle of the vehicle relative to the rail,and timely transmit the detection results to the cab,so that the driver can adjust the vehicle position to make the wheelset drop on the track accurately and efficiently.Based on the literature review of vehicle assisted driving and vision measurement technology,this paper analyzed the actual demand of road-rail vehicle for the wheelset drop guidance,and determined the overall design scheme of intelligent wheelset drop guidance system based on machine vision.The whole system includes two parts: hardware and software.This paper completed the selection of hardware platform and field assembly debugging,and designed the software system based on Lab VIEW.The software system was divided into lower computer software system and upper computer display system.The former mainly realized the functions of image acquisition,image processing,target recognition,analysis and calculation,and data transmission;the latter was mainly responsible for data reception and display.The wireless data transmission between upper computer and lower computer was realized based on TCP/IP protocol.The paper studied the method of realizing the function of wheelset drop guidance by detecting the wheelset drop marking line beside the rail.Firstly,this paper introduced the implementation process of camera calibration and the preprocessing method of marking line image,and then studied the edge detection of marking line and key marking line recognition algorithms.When performing marking line recognition,this paper analyzed and compared two kinds of marking line recognition algorithms based on Hough transform and least square method.The analysis results showed that the marking line recognition algorithm based on least square method was more stable.Finally,the ranging model based on the marking line image is constructed,and the measurement and calculation of related data are completed.Considering that the marking line may be contaminated or damaged after long-term use,which may lead to the failure of the wheelset drop guidance system based on marking line detection.In order to improve the adaptability of the system,this paper further discussed a method of directly detecting the rail to realize the function of wheelset drop guidance.Compared with the marking line,the rail is less recognizable in the image than the marking line and has some problems such as unobvious edge features.For the reason,the paper studied the corresponding rail image preprocessing and threshold segmentation algorithm,and adopted a rail recognition algorithm based on the connected region feature,which can accurately identify the rail region under complex conditions,and then realized the measurement and calculation of related data.Finally,in a certain conversion operation section of the road-rail vehicle,the installation,debugging and testing of equipment and system were carried out.The test results showed that the distance measurement error is within ±2mm and the response time of the system is about140 ms when detecting based on the marking line image;the distance measurement error is within ±5mm and the response time is about 240 ms when detecting based on the rail image.Both methods met the expected technical requirements and proved the feasibility and effectiveness of the system. |
