Lane Recognition Method Based On The Direction Of The Tunable Filter

Lane detection is an important component of intelligent vehicles vision navigation systems that require information of lanes to decide driving routes of vehicles. This paper researches the problem of lane detection in structured road images and does some work on how to extract lane information in road images efficiently, including image enhancement, extraction of lane-marking feature and lane-marking detection.1) Several kinds of commonly image enhancement algorithms with small complexity and small numeration are analyzed and by the requirements of road image processing, the appropriate method is chosen.2) Several conventional algorithms for edge extraction and steerable filters algorithm are compared. Comparison shows that the method can intensify lane information effectively and eliminate image noise.3) The algorithm for lane detection is presented in details, including initial lane detection and fitting lane model. During the initial detection, road images by CCD are firstly processed by the image enhancement algorithm and the lane-marking extraction algorithm which are determined above. Next the processed images are divided into near vision field and far vision field, and then the edge distribution function (EDF) and the Hough Transform are used to determine lane direction. Last based on the analysis of lane direction characteristics in images, region of interest (ROI) are established. During the lane fitting phase, a linear-parabolic model is firstly constructed to approximate lane boundaries: the linear part of the model is used in the near vision field, while the parabolic model in the far field. Then least squares method is used to fit lane edge pixels.Based on above studies,the paper proposes an algorithm for lane detection based on steerable filters. Simulations under different road conditions show that the lane detection algorithm can detect lanes stably.