Moving Human Target Detection and Tracking in Video Frames

摘要

The conventional method for moving human target detection and tracking has come across a major setback due to various hindering factors such as environmental lighting conditions, temperature, etc. Similarly, it has been noticed that the manual selection of moving human targets in a video sequence does not provide convincing results either. In this paper, a new method for moving human target detection and tracking is proposed. It involves two stages. The first stage consists in the detection of moving human targets and the second one in target tracking based on the Continuously Adaptive Mean-Shift (CAMShift) algorithm. In the first stage, in order to select the moving target, the background subtraction method and frame subtraction method are combined. The Region Of Interest (ROI), which is usually the moving target is identified. In the second stage, target tracking is performed by choosing a centroid pixel point over the ROI, which is then used by the CAMShift algorithm. The proposed method has shown outperforming results for various performance parameters such as precision, accuracy, recall, and the F1-score under three different lighting conditions. The results obtained also show a reduction in time complexity in comparison with the state-of-the-art algorithms.