A behavior tracking algorithm based on gradient vector flow and linear discriminant analysis

Abstract

Aiming at the problem of scarcity in traditional tracking algorithms in order to track behavior more rapidly and accurately, a behavior tracking algorithm based on linear discriminant analysis combined with the gradient vector flow is presented. First, the gradient vector flow is calculated in the reference coordinate system, the appropriate curve parameters are set, and the position of the behavior is determined. Then, the discreteness is calculated within the range of behavior, the discriminant function is defined, and the maximum value and the best eigenvector are obtained. Last, the behavior is detected when the value of the dispersion degree is less than the set threshold. The new algorithm is not affected by background changes. Compared with existing algorithms, the simulation results show that the method is easy to realize and has a more stable tracking effect.