Target tracking method based on local features and scale pool

Abstract

The invention discloses a target tracking method based on local features and a scale pool, belongs to the technical field of grayscale image target tracking, and solves the problem of poor target tracking algorithm accuracy in complex environments such as illumination change, scale change and background interference. The method comprises the steps of a first frame image obtaining a target according to initial information, training a classifier on the basis of two features of the target, obtaining and then initializing a target model and a classifier regression coefficient, a second frame imageobtaining targets of different scales through a scale pool, and extracting a first feature and a second feature; based on the initialized target model and the classifier regression coefficient, obtaining a multi-layer kernel correlation filtering response graph of the two features; and performing linear interpolation to the same size, performing weighted fusion to obtain a multi-layer kernel correlation filtering response graph, further obtaining a prediction position and a prediction scale of the target, namely finishing one-time target tracking, and if the tracking is not finished, realizing tracking from the second frame of image to the third frame of image until the tracking is circulated to the last frame of image. The method is used for target tracking.

Description

Technical field [0001] A target tracking method based on local features and scale pool, used for target tracking, belongs to the field of gray image target tracking technology. Background technique [0002] Target tracking has very important significance and value in the field of computer vision. It has wide applications in many fields, such as intelligent video surveillance, medical care, human-computer interaction and other civilian fields. In the military, it can quickly and accurately target enemy moving targets. Search and tracking etc. Target tracking is mainly divided into generative model and discriminative model. The generative model establishes the target mathematical model to complete the matching between the candidate target and the target model, and takes the most similar candidate area as the prediction target. The discriminant model is composed of a training set composed of positive samples belonging to the target and negative samples belonging to the background ...

AI Technical Summary

Problems solved by technology

[0005] In view of the above-mentioned research problems, the purpose of the present invention is to provide a target tracking method based on local features and scale pools to solve the problem of low robustness and poor accuracy of target tracking algorithms in complex environments such as illumination changes, scale changes, and background interference. question

Images