The advancements of deep learning in computer vision has resulted in intelligent systems, which are capable of making appropriate decisions based on different situations. The aim of this project is to develop an intelligent system to automatically analyze video acquired from surveillance cameras. Visual object tracking is one of the major components in developing such systems. The object tracking predicts the trajectory of moving objects across the video frames. These predicted trajectories could further be helpful to solve the multiple tasks such as anomaly detection including red-light jumping, over-speeding, theft-detection. Therefore, the precise localization of target objects is very crucial for analyzing surveillance videos. Multiple challenges such as occlusion, weather conditions, motion blur limit the capabilities of existing tracking systems. The objective of this PhD project is to propose a tracking system, which overcomes these limitations by combining multiple modalities such as RGB, Depth, Thermal, etc.
This PhD project is directly funded as part of the Milestone Research Programme at AAU (MRPA).