Tracking dim-small object based on the hyperspectral features

In order to track the dim-small object in fast moving scenario, a precise tracking method based on the hyperspectral features is proposed since the traditional full color tracking seems impossible for unobvious color and contour features. A multi-dimensional feature space is extracted with the spectral fingerprint model. To track the dim-small object with high speed, this paper integrates a Kalman filter into the nonparametric kernel density estimator which is built with the probability histogram of spectral features. To avoid the object jump incident, a layered particle filter is introduced into spectral tracking algorithm. The experimental study and analysis show that the tracking algorithm based on the hyperspectral features is accurate, real-time and robust.