Identification Of Lane Boundary Under Different Lighting Based On DS Evidence Theory

Detecting lane boundary is the most important problem in the study of Intelligent Vehicle and Driving Safety Assistance System. At present the available detecting method based on monocular machine vision can identify bend path on daytime under complex condition, and it has higher reliability and real-time quality compared with traditional identification method. While it still has three disadvantages. First gray feature is selected as global feature which is vulnerable under different lighting conditions. Second lane direction feature is not selected, and moreover, gray and grad features have just 4 kinds. Third objective function is adopted linear classifier which has simple form and easy to achieve but the reliability should be more improved.Lane boundary identification method based on DS evidence theory is advanced in this paper under different lighting conditions. First the lane boundary features of grad, gray and direction are extracted in local areas, and the features are hardly interfered by variation of global gray with different lighting and noise of shadows, furthermore, gradient and gray features are broadened to 14 kinds. Second features of lane boundary and the proportion of effective statistical units are fused by DS evidence theory, and identification quality is fitted by objective function accurately. Third the ranges of test thresholds are set at different lighting conditions. Forth the ant colony algorithm is applied in lane identification, and with the guidance of the objective function designed the positive feed back of hormone identify lane boundary as soon as possible.Lane boundary identification system is based on TMS320DM642. To improve real-time capability, the code optimizing methods, which include optimization of application layer, intrinsic function and packaging data, are used to optimizing process. It provides a good hardware platform for industrialization system. The method of lane boundary identification system under different lighting conditions is installed in experiment vehicle and carries out road tests. The tests show that the method can identify the lane boundary accurately, and have real-time capability.