Moving Target Detection And Tracking In Road Vehicle Vision Navigation

Real-time detecting and tracking moving targets leading ahead is very important for road vehicles to avoid collision, and it is one of the most important modules of vision navigation. This paper researches on the problem of detecting and tracking a moving target leading ahead on own lane under the highway or level road environment, by processing and analyzing a long sequence of images which are gathered by a CCD camera mounted on a driving car outdoor. The feature-based target detection algorithm and the model-based 3D-motion parameter estimation algorithm fulfill the detection and tracking of moving target. The core of our algorithm is to detect a target by finding the horizontal and vertical edge, and to verify whether a target exists or not by computing the local rear symmetry of the target in the image. Building the state model that can reflect the relative motion between the target and the camera and the target geometry parameters, the perspective model that can reflect the features of the target and the road, then estimate the 3D-motion parameters by generalized nonlinear Kalman filter. In the process of Kalmsn filtering, our algorithm uses the sequential process and the U-D covariance factorization technique to improve the computing speed and numeric stability of filtering. This algorithm is fast, robust and practical. At the same time, the experimental results that prove that our algorithm is satisfying are given in this paper. So the approach we adopt here is a feasible solution to solve the problem of detecting and tracking targets in the road vehicle vision navigation. Feng Lijun [Application of computerl Directed by Prof. Su Kai_na...