Abstract:

In order to build a right focus window to cut down the compute time and overcome the disturbance of the background for the photoelectrical theodolite in range measurement, a novel method is proposed. Firstly, the method regards the target’s centroid gotten by the Meanshift tracking algorithm as the reference point to compute the adscription of pixel. Secondly, visual saliency weighted by the adscription is detected by Log-Gabor filters and the Gaussian difference model to ascertain the target area. Then, the adaptive kernel function is designed to track the target combined with the Epanechnikov function, which can reduce the effects of fixed track window and background pixels. Finally, the focus window is built with boundary extension by the estimating diagonal area, and the size and position of the focus window are self-adaptive with the target. After plenty of experiments on the photoelectrical theodolite, the results show that the proposed method can effectively track moving targets in real time and build focus window exactly. To deal with image of 720 pixel×576 pixel, the compute time is less than 20 ms. The proposed method satisfies the practical demand of the photoelectrical theodolite and is of great value for engineering application.