Design And Implementation Of Lane Marking Detection And Tracking Algorithms Based On Particle Filting

The primary coverage of the ITS (Intelligent Transportation System) is the detection of the driving environment. Hereinto, lane detection is one of the key technologies of ITS. Lane marking is regarded as the basic constraint of the lane. And it can show the position of the vehicle in the road, as well as the direction in which the vehicle is heading, and it also can provide the information of the road surface. So the detection and tracking of lane marking based on vision is requisite. Presently, most existing algorithms of lane detection and tracking are used in freeway and not designed to handle road singularities such as freeway exits, double painted lines, etc. And they are based on various additional assumptions, such as the vehicle is initially located in a straight portion of the lane, there are no linear structures in the image, except for the lane marking. It is difficult to approach such initial condition.This paper firstly analyzes the current situation of lane marking detection and tracking algorithms based on visual system. On ground that comparison of various algorithm, the existent issues of the algorithm are expounded. Secondly, a lane marking detection algorithms based on Confidence Measures method is proposed. Edge points cue is used to detect the lane marking and a road orientation estimation method is used to delete the edge lines which are impossible attribute to lane markings. And a finite-state machine which is fused with multi-frame detected results decides the state of a lane marking. It can terminate the algorithm of disappearing lane marking in time. Afterwards, in light of Particle filter theory, an algorithm of lane marking tracking based on CONDENSATION algorithm is introduced. According to the priori knowledge of the lane marking, the state evolution model and the observation model are builded and the description of the posterior density is also obtained. Moreover, partitioned sampling method is used to enhance the performance. It makes the algorithm accurate and reliable. At last, the lane marking detection and tracking algorithm is evaluated in different video sequences, including illumination, weather and road, demonstrates its effectiveness and robustness. The algorithm runs in real-time at rates of about 30 Hz.