Multiple camera based multiple object tracking under occlusion: A survey

In recent years, vision based technologies have gained immense attention across academia-industries to enable optimal surveillance solution for event monitoring, analysis and control. However, the complexities of real time environment and expected functional characteristics often put question over existing approaches and their efficacy. In this paper, a number of the existing approaches for vision based object detection and tracking have been discussed. One of the important issues which come across in the object tracking is Occlusion. However, a very few efforts have been done to alleviate occlusion issues so as to enable optimal moving object detection and tracking. This review paper reveals that to enable efficient multiple object detection and tracking under occlusion conditions, the combination of different features such as depth data, geometry, textural, color feature metrics, speed, etc can be taken into consideration. With multiple synchronized cameras set up, 3D techniques including depth estimation, azimuth, texture, color, geometric information and speed etc, in conjunction with optimal trajectory estimation model can be vital for multiple objects tracking under occlusions. In addition, efficient region of interest (ROI) feature extraction, trajectory estimation and feature fusion across synchronized cameras can be explored to track multiple objects under occlusions. Synchronizing multiple cameras with best feature training, association, continuous object skeletal tracking using stochastic trajectory prediction and classification; can be an effective solution, especially for traffic surveillance or outdoor surveillance applications.