Extended target stable tracking point extraction method based on local and global coupling

Abstract

The invention discloses an extended target stable tracking point extraction method based on local and global coupling. According to the method, preprocessing such as denoising and enhancement is performed on an image to be processed; a GMS (Grid-based motion statistics) algorithm is used to perform feature point matching on two adjacent frames of images; generalized Hough transformation is performed on matched feature points, and features that contribute to the peak points of an accumulator are obtained and are adopted as basic feature points; the homography mapping matrix of the feature points of the two adjacent frames of images is obtained by using the basic feature points; and a feature point of sub-pixel precision of a current frame of image which is obtained by means of mapping on the basis of a tracking point of a previous frame of image is calculated through using the homography mapping matrix, and the feature point is adopted as the tracking point of the current frame of image. The tracking point obtained by using the method of the invention is a virtual global feature point which is obtained by means of mapping on the basis of a plurality of local matched feature points on a target; and a condition that the target always has stable and obvious features is not required. The method has high robustness to the translation, scale, illumination, blur, attitude and self-occlusion of the target, and even changes such as the escape of a part of the region of the target from a field of view.

Description

technical field [0001] The invention belongs to the fields of image processing, computer vision and target tracking, in particular to a method for extracting extended target stable tracking points based on local and global coupling, which is a local and global method combining GMS algorithm, generalized Hough transform and homography matrix mapping A Coupled Extended Target Stable Tracking Point Extraction Method. Background technique [0002] In imaging and tracking systems, in order to improve the stability and accuracy of target tracking, a smaller field of view is generally used, causing the target to occupy more than 1 / 3 of the area in the field of view, and may even appear in an extended form beyond the field of view. The posture of the extended target changes significantly during the movement process and some areas leave the field of view. At the same time, atmospheric turbulence and system jitter will cause translation and blur problems of the target. In addition, l...

AI Technical Summary

Problems solved by technology

These methods can achieve a stable tracking effect when the target has obvious and unchanged features all the time, but when the target is partially occluded, part of the area leaves the field of view, or is severely rotated due to attitude changes, the selected features are no longer obvious or even When invisible, there will be jitter and drift or even tracking failure

At the same time, illumination changes, scale changes, blurring, etc. may also cause the original tracking feature points to fail.

In response to these situations, some researchers proposed to use the generalized Hough transform ([4]Ballard D H.Generalizing the Hough transform to detect arbitrary shapes[J].Pattern Recognition,1981,13(2):111-122.) to extract feature points Track extended target ([5] Hu J, Peng X. High precision localization for profile dextended object based on GHT [J]. Optik-International Journal for Light and Electron Optics, 2014, 125(3): 1297-1302.), in the target When some areas leave the field of view, this method can still lock the position of the original tracking point, but when the scale and attitude of the target change, and the outer contour changes greatly, the tracking point will drift

Images