Moving targets detection and tracking based on improved codebook algorithm and Kalman filtering
Wendan Su, Huiping Zhuang, Xiaohong Qiu · 2017
Multiple moving targets detection and tracking in a video is one of the most important research fields of the computer vision. Considering the scenario when different objects are connected due to the shade during multiple moving targets detection, this paper proposes a modified background subtraction to detect objects in real-time with the shade removed. To begin with, a background model is built to remove the shade using maximum chromaticity differential method. The background is updated in real time so it can easily adapt to the changing environments (e.g., sunlight). In addition, Kalman filters is applied to a feature, which is extracted from the centroid of the moving target by means of the morphology method, for tracking the moving target in the camera in real-time. By the calibration of the camera, the coordinate of the picture is converted into the world one such that the moving target can be tracked in real world. The conducted experiment has shown good performance, thereby presenting validity of the proposed method.