A new approach of vehicle detection in complex environments

Vehicle detection is a major way of subtracting moving vehicles from traffic flow images in ITS. In order to acquire an accurate background from the constantly changing traffic environment, we design and realize a robust background updating model based on single Gaussian distribution. Then, we use the method of local normalization to subtract the texture image, and carry out texture segmentation to acquire the moving vehicles by virtue of the Laws´ texture energy method. The experiment shows that this method can efficiently eliminate the effect of the vehicle shadows and bright pavements and it is especially suitable for the nighttime vehicle detection in complex traffic environments with streetlights.