Unsupervised video multi-target tracking based on fast resampling particle filter

In order to improve non-supervision and monitoring effect for moving target of video, target tracking algorithm without supervision video based on resampling particle filter is proposed. A fast resampling particle filter algorithm designed in the paper is adopted, and it is based on quasi-Monte Carlo method which owns the property of low difference for determining sequences for acquiring more uniform sample distribution in space and is able to avoid degeneracy of sampling particle for improving calculation efficiency and accuracy of particle filter algorithm; subsequently, background difference algorithm is used for reducing algorithm for frame of video supervision image. Dynamic particle filter is established according to quantity of moving target in image with color distribution character to trace it, and the establishment of filter and reducing algorithm is used for reducing algorithm redundancy; eventually, according to simulation test in instance of supervision image, it shows that target tracking accuracy and calculation efficiency of the proposed method are better.