Research on Small Object Detection and Tracking Based on Particle Filter
Zhenhua Wei, Yanping Liu · 2009
The problem of tracking small objects poses a number of challenges, which dues to the ambiguity of the observations and the presence of partial or complete occlusions. In this paper an efficient method of tracking before detecting for small object is proposed in the low signal to noise ratio image sequences under complex background. In the tracking phase, we introduce auxiliary particle filter to obtain particle collection, which is according to noise sample set. In the detecting phase, we make use of target trajectory search and make a decision based on energy accumulation. We use detection based on neighborhood decision algorithm. In order to reduce the detection of reproducibility and false, a method of special centroid extraction based on the small target is proposed. The experimental results show that this method is able to obtain trajectories of the target rapidly and significantly increase the detection efficiency based on the running trajectories.