Research On Correlation Tracking Algorithms

Correlation tracking algorithm is being used in many tracking technology of sequence infrared images, the existing algorithms need compute a large quantity, as well as the target trajectory requiring prediction and the target being blocked during tracking make the correlation tracking algorithm difficult to meet the real-time and stability requirements. This paper researches the sequential similarity detection algorithm (SSDA) and the maximum neighbor distance (MCD) algorithm which both use image-based features. This paper selects Canny operator which has obvious detected effect to check the edge of the template image of the target via simulation comparison of several common edge detectors, which let edge pixels participate in correlative calculation to reduce the computational complexity of adaptive threshold SSDA; for MCD algorithm, via simulation comparison of linear prediction, square prediction and integrated prediction, we select a integrated predictor which predicts better to reduce the search area and improve the tracking stability.Simulation results showed that the computational quantity of the improved SSDA algorithm reduced a lot, the improved MCD algorithm can track stably even though the whole target or part of target was blocked, the stability and anti-jamming of MCD were largely improved.