Vehicle Contour Tracking Based on Kernel Histogram and Level Set

摘要

In order to extract the vehicle contour accurately, the paper proposes the track method of vehicle contour combining the kernel histogram and the active contour tracking method, which has comprehensive advantages. Firstly, the vehicle model is described in RGB color space. The series quantization problems are solved by the N-Bin histogram algorithm. The fixed window has the tracking errors in the traditional Mean Shift algorithm. This problem is solved with image amount information. Then the vehicle initial area is located from the Mean Shift algorithm combining the improved GM(1, 1) forecasting model. Finally the vehicle contour has a further evolution with level set method based on the Mumford-Shah model to get a more accurate vehicle contour. The experimental results show that the proposed method has a more tracking accuracy result comparing with the traditional algorithm based on kernel.