A Moving Human Detection and Tracking using Combination of HOG and Color Histogram

Detection and tracking of a moving target in a crowded area using camera surveillance is still an open topic that to be resolved. Determining a single target instance known among other objects in the room becomes a problem. We propose a color-based human tracking algorithm by considering the correlation distance from the color histogram used as the target selection. Furthermore, we also create a new dataset based on the INRIA dataset to minimize the detection area. Then, we apply the HOG extraction feature to the human detection system. We apply the Recall-Precision curve to evaluate the performance. The accuracy and precision rate has increased by 10 times. Experimental results are also provided in performance metrics to prove the reliability of our proposed method.