Camshift Target Tracking Based on Gaussian Background Difference and Adaboost Combination

Aiming at the problem of the traditional AdaBoost detection, such as false detection and large amount of calculation, this paper proposes a new detection method combining Gaussian background difference and AdaBoost. First, we use Gaussian background difference and morphological processing to remove unwanted information and extract the motion area. Then we use AdaBoost algorithm to identify the target of the moving area, determine the exact location of the vehicle and record this position, so that the vehicle can still be detected when it is decelerated or stopped. Set CamShift’s initial search window to a detected vehicle position and then track the vehicle’s real time position by iteratively approximating the exact position of the target. The experimental results show that this method can achieve real-time detection and tracking of vehicles with complex background and has good robustness.