Research On Key Techniques Of Advanced Driver Assistance System Based On Machine Vision

Advanced Driver Assistance System is an important part of Intelligent Transport Systems (ITS), which has significant meaning to reduce traffic accidents, reduce people injury and improve the capacity of road transportation. Machine vision technology is widely used in different kinds of Advanced Driver Assistance Systems because of its advantages of similar to human vision and low cost. In order to provide more accurate and timely road environment information, many key technical issues, such as lane detection, vehicle detection and recognition and traffic sign recognition, are deeply discussed in this paper. Then some solutions based on machine vision are proposed. The main research contents and achievements are summarized as follows:(1) The intrinsic and extrinsic parameters of the onboard camera calibration method is deeply discussed. The planar pattern method is used to calibrate the intrinsic parameters. In order to avoid the complicate calibration and improve the calibration precision, the extrinsic parameters can be real-time adjusted according to the position of vanishing point and the relation of the slope of the both lanes.(2) A new lane detection algorithm based on bi-directional sliding window technology is presented. Firstly, a new method, which integrated EDF and Hough transform, is proposed to obtain the linear part of the lanes. Secondly, the bi-directional sliding window technique is applied to extract the real lane line feature points. Finally, a linear-hyperbolic lane model is used to depict the shape of the lanes.(3) The issues about the front vehicle detection and recognition are deeply discussed. The candidate image regions of the vehicles are firstly extracted by using the lane constrain and the gray statistic information of the roadway. Based on the vertical symmetry characteristics of vehicle images, a vertical symmetrical histograms of oriented gradients (VS-HOG) descriptor is proposed creatively for extracting the image features. In the classification stage, an extreme learning machine is used to improve the real-time performance. Compared with other classical methods, the vehicle verification algorithm based on VS-HOG and ELM achieves a better trade-off between cost and performance.(4) A novel system for the automatic detection and recognition of traffic signs is presented. The system consists of two stages:detection and recognition. In the detection stage, candidate regions are detected by using the color features of the pixel and the Multi-scale sliding window technique. In the recognition stage, cascaded SVM classifiers are utilized to recognize the shape and content of the traffic signs. The experimental results show a high performance both in accuracy and efficiency when using the system to detect and recognize the traffic signs.