Compensation algorithm for optical flow vectors based on image sequences of the spot

Abstract

A key problem in the processing of image sequences is estimating the motion between video frames, often referred to as optical flow estimation. For the problem of the target spots loss in image sequences when aircraft's 3D attitude parameters is measured based on optical measurement, an effective optical flow vectors compensation algorithm from multi-frames image is proposed. First, the optical flow vectors between frames of spot target image sequences are calculated by Lucas-Kanade algorithm, and then choose the optical flow vectors nearby the images of lost spot target. The chosen vectors are used for estimating spot image. The compensation model of optical flow vectors is set. Then, the lost spot target center is got by using inverse distance weighted averaging and the weight-value is the difference of the frame count between real image and estimated image. By experiments analysis, the proper calculating scope of this arithmetic under the expected precision is determined. This algorithm compensates the calculation errors caused by the lost target in image sequences efficiently.