A Novel Approach for Lane Detection Sensor via Maximum a Posteriori Estimation and Inertia Narrowed Region Particle Swarm Optimization

摘要

A novel approach for lane detection sensor via maximum a posteriori (MAP) probability estimation and inertia narrowed region particle swarm optimization (INR-PSO) is proposed. The real lane model is studied via analyzing the relationship between the world and pixel coordinates, and then an appropriate lane model in the image is deduced. Based on the MAP estimation, suitable prior probability distribution function and likelihood function are proposed to extract the lane parameters. In order to obtain the result as fast as possible, inertia narrowed region particle swarm optimization (INR-PSO) is proposed and employed to accelerate the calculation of lane parameters. Experimental results under a variety of situations reveal that the proposed approach can extract the lane sensor parameters exactly and quickly, which demonstrates that the method is efficient.