The Research For The Height Of Barriers Based On Vision Measurement While The Vehicle Is Driving

The thesis for research vehicle based on vision measurement the height of barriers,automatic avoidance of the vehicle, as well as safe driving conditions complicated vehicle provide important information, and for the development of vehicle safety driving visual auxiliary system provides the research foundation. In the automotive driver assistance systems, accurate detection of height information has very important significance, in the vehicle. The thesis makes a preliminary study on vision measurement of the barrier height.firstly, use the camera to capture images of the barrier, and then by setting appropriate threshold, using the threshold segmentation method to segment the barrier area, and extract the characteristics of its angular point, finally, establish the mapping relationship between the image height(the barrier vertex rather than on the ground projection point of the difference of the image ordinate) and the actual height of the barrier, it is to build on-board camera calibration model. Using this model to measure the actual height of barriers on the road, the main research content is as follows:(1) Image segmentation and feature point selection, by using the method of threshold segmentation to extract the image of obstacle. According to the characteristics of the barrier is an independent body shape in the image, extraction of feature points. Because the thesis is the study for the barrier height of vision measurement, so the coordinate difference between the barrier vertex and the projected point of image coordinates as the research focus.(2) Feature extraction and barrier recognition. The image of the height and width of the barrier area of characteristic is analyzed, by the target area in the image barriers information indirectly to express obstacle of actual height and width. Choose the appropriate recognition methods, based on the barrier height and width to classify barriers on the road. In this thesis, the recognition method based on BP neural network, with the characteristic of the barrier for the input, with the type of barriers for the output, design a suitable classifier for identification.(3) Under binocular vision system, select the appropriate algorithm to establish the mapping relationship between the image coordinate and the actual location coordinate, to realize the mapping relationship between point and point, explored to measure the position of scene. Analyze the vertex of barrier, based on the idea of using object vertex express height, study obstacle height. And then study the mapping relation between the image height and the actual height of scene under the binocular vision system, establish themapping relationship between line and line, achieve the purpose of measuring scenery height. In this thesis, through the combined method of interpolation method and the least square method, establish the mapping relationship between image height and the actual height of barriers, namely establish a new model of on-board camera calibration.(4) The vision measurement of the barrier’s height while the vehicle is moving. Use the image information on the step calibration of binocular camera acquisition obstacles on the road, and process the obstacle image, extract obstacle image height as the on-board camera calibration model input, the output of the model is the obstacle of the height of the measurement. Experience the maximum error of the obstacle between the actual height and the measured height is 0.04 dm, the average error rate is 1.32%, and the average precision can be up to 98.68%.