Unmanned aerial vehicle target tracking method based on anchor frame matching and Siamese network

Abstract

The invention discloses an unmanned aerial vehicle target tracking method based on anchor frame matching and a Siamese network, and the method comprises the following steps: building a five-layer Siamese network and a region suggestion network RPN comprising a classification branch and a regression branch, obtaining a target position through the classification branch, and obtaining a target scale through the regression branch; applying a dynamic anchor frame matching criterion to a training stage and for optimizing a classification task and a regression task, so that prediction of one task can dynamically design an anchor frame sample to improve a model of the other task, two branch models of RPN can learn each other, and the position and the scale of a target can be obtained more quickly and accurately; in the tracking stage, obtaining first K candidate frames with the highest score near a target to establish a target search library, then finding a most reliable prediction frame out by using a region of interest (ROI) perception model, so that the influence of a complex background on a tracking algorithm is reduced. According to the method, the precision of a tracking algorithm is effectively improved, and good robustness is achieved.

Description

technical field [0001] The present invention relates to a video target tracking method, which integrates advanced technologies in many fields such as image processing, pattern recognition, artificial intelligence, automatic control and computers, and particularly relates to an unmanned aerial vehicle video target tracking algorithm for anchor frame matching and Siamese network . Background technique [0002] Visual object tracking has been applied in many fields, especially in the field of unmanned aerial vehicles. It has been widely used in aerial reconnaissance of drones, aerial aircraft tracking and aerial refueling. Due to the continuous change of the drone's shooting angle and flight speed, there will be complex situations such as similar objects around the target and occlusion. Therefore, it is of great significance to design an effective and robust UAV target tracking algorithm. In recent years, trackers based on the Siamese network have good tracking accuracy and ...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the problems that targets in UAV video are easily occluded and similar objects interfere, the present invention designs a UAV target tracking algorithm based on anchor frame matching and Siamese network, combined with dynamic anchor frame matching strategy, Siamese network and area suggestion network A Siamese network model is constructed, through which more robust feature information of the target can be obtained

Images