Study Of Target Tracking Algorithms In Video Image Sequences

Target tracking of image sequences is a process to locate and track a target in each image frame. At first, lock the target to be tracked, and then take appropriate response measures to target according to actual needs. On the basis of study and simulation of many kinds of tracking algorithm, we put the emphasis on two kinds of target tracking algorithm, seeking for new solution to different question existing in the tracking algorithms.This paper has studied the gray template related matching track algorithm in time domain and phase correlation algorithm in frequency domain on the gray image’s simple target tracking question in complex background. To solve the problem of target’s size and pose changing rapidly, we combine Weight Average-removed Normalized Cross Correlation (WNCC) and Fourier-Mellin transform based on all phase spectrum analysis (apFMT) to track targets. The algorithm has great performance of rotation invariance, scale invariance and anti-noise. Simulation results show that this algorithm has a good effect, and it can get stable tracking.Artificial neural network has the characteristic of non-linear large-scale parallel distributed processing and great ability to handle uncertain information. If it is used in image matching, not only can improve real-time, but also greatly reduce the requirements in image quality, which can improve the performance of the tracking system. Combined PCA algorithm with BP neural network, we design an image classification discriminator and mention a new object tracking algorithm based on BP neural network. Simulation results show that this algorithm has better robustness and a good effect in object tracking with complex background.