Optical remote sensing video object tracking method based on depth SR-KCF filter

Abstract

The invention provides an optical remote sensing video object tracking method based on depth SR-KCF filter, is used for improving the tracking accuracy on the basis of ensuring the tracking speed. Therealization steps are as follows: obtaining the training set of the object detection network R-FCN; training target detection network; optimizing the target tracking algorithm KCF to obtain SR-KCF; setting The judging condition of target state in target tracking algorithm SR- KCF; Constructing detection-assisted tracking algorithm; setting matching condition between tracking target and detectiontarget; Acquiring information of an object contained in a first frame image of an optical remote sensing video; acquiring information of an object contained in the second frame and subsequent images of the optical remote sensing video; outputting optical remote sensing video object tracking results. The invention can extract more advanced features of the target by using depth learning, adds a background suppression factor and a sparse response factor to the correlation filtering algorithm KCF, and effectively improves the tracking accuracy.

Description

technical field [0001] The invention belongs to the technical field of remote sensing video processing, and relates to an optical remote sensing video target tracking method, in particular to an optical remote sensing video target tracking method based on deep SR-KCF filtering. It can be used in land surveying and mapping, mineral resource development, smart city construction, traffic facility monitoring, agricultural yield estimation, forestry resources survey, ecological environment monitoring, disaster prevention and mitigation, etc. Background technique [0002] Object tracking refers to predicting the size and location of objects in subsequent frames given the size and location of objects in the initial frame of a video sequence. According to different imaging methods, target tracking can be divided into optical video target tracking and remote sensing video target tracking. Remote sensing video target tracking includes infrared video target tracking and optical remote...

AI Technical Summary

Problems solved by technology

However, its disadvantages are that, on the one hand, this method uses HOG or grayscale to extract the features of the target, which can only extract the rough features of the target, which is easy to cause the problem of tracking delay, which in turn affects the tracking accuracy. The KCF algorithm ignores the factors that the tracked target may be blurred or partially occluded when designing the objective function. When the target is blurred or partially occluded, it is easy to cause target tracking loss, and its tracking accuracy is more limited.

Images