Font Size: a A A

Design And Implementation Of An Algorithm To Recognize The Rail Ahead Of The Train

Posted on:2018-05-22Degree:MasterType:Thesis
Country:ChinaCandidate:Y Z WangFull Text:PDF
GTID:2322330515959784Subject:Computer Science and Technology
Abstract/Summary:PDF Full Text Request
In recent years,with the rapid development of urban rail transit,subway,light rail and other train traffic safety becomes increasingly important.The obstacle detection system improves the driving safety of the train by assisting the driver to check the track in front of the train.As for the obstacle detection system,it is vital to recognize the rail ahead of the train efficiently and accurately.In this paper,an approach is proposed to recognize the rail from image taken by train-mounted camera.The recognition process is mainly divided into two steps.Firstly,recognize the rail near the train.Secondly,based on the result of first step,the improved method of seed region growth is used to recognize the distant rail.As for the near rail recognition,this paper uses the existing rail recognition algorithm based on curvature mapping image and make further improvement.The improvement is mainly reflected in the creation of curvature mapping image.The creation process in this paper is more convenient,and the curvature mapping image created is more precise.Besides,this paper also make some improvement on selecting the closest matching curvature according to the gradient of input image and the curvature mapping image.As for the distant rail recognition,the method used in this paper is based on local gradient information.Firstly,a metric is designed to measure the similarity between the image in a region and an actual rail image,according to local gradient information.Secondly,select seeds from the near rail,including position and direction.Thirdly,find the best seed(mainly depended on the similarity in first step)in the neighborhood of current seed.Fourthly,according to the best seed's position and direction,extend to next seed.Finally,repeat third and fourth step until no seed satisfying the constraint between the left and right rails can be found,when the distant rail has been recognized.Moreover,this paper proposes a method to calibrate camera's external parameters using two parallel lines with known distance.As the straight rail can be taken as two parallel lines,the calibration procedure is rapid and convenient.
Keywords/Search Tags:intelligent train, rail recognition, image gradient, seed region growth, camera calibration
PDF Full Text Request
Related items