Multi-object tracking system using dissimilar apparatus in video sequence

A combination approach using vision and non-vision data concurrently is to perform multi-object tracking in video sequence. Before applying modified matching stage from Gale-Shapley algorithm the two types of data will be refined to pair in advance. The confirmation of object IDs and their positions is accompanied by the trajectory analysis of each object for error correction during the matching step. The proposed method indicates the consistency and the accuracy of the system in a long video sequence. While the non-visual data keeps the IDs of multiple objects regularly when the objects are undetected or mis-detected using the visual data, the visual data corrects the technical position error of non-visual data. Due to this interaction we expect that the proposed system will be strengthened for visual object tracking. Future work will aim to capitalize on the tracking ability of proposed system with other positioning devices.