Mobile robot infrared target tracking method and system based on unsupervised optical flow network

Abstract

The invention provides a mobile robot infrared target tracking method and system based on an unsupervised optical flow network. According to the mobile robot infrared target tracking method, an unsupervised optical flow network which can achieve the end-to-end training is used for extracting optical flow information of a previous frame, and the optical flow information is used for aligning featuremaps of a plurality of previous frames. The first several frames of features are fused through a spatial and temporal attention mechanism to obtain a feature map of the target, and finally a trackingresult is obtained by using a correlation filter according to the feature map and the feature map of the frame to be predicted. The beneficial effects of the invention are that the unsupervised optical flow network which can achieve the end-to-end training is used, the optical flow features are extracted, the features of a plurality of frames are fused, and the tracking effect is improved. Especially, targets moving quickly frequently occur in the tracking process of the mobile robot can be effectively tracked by using the method.

Description

technical field [0001] The invention relates to the technical field of visual target tracking, in particular to an infrared target tracking method and system for a mobile robot based on an unsupervised optical flow network. Background technique [0002] Visual perception is an important and indispensable part of the perception system of intelligent robots. Visual object tracking is the supporting technology of visual perception. The robot first needs to position and track the target before it can interact. Visual object tracking technology is a research hotspot in the field of intelligent robots, and is widely used in robot visual tracking and navigation, intelligent monitoring and other directions. The task of visual target tracking is to give the position and size of the target to be tracked in the initial frame of the video, and predict the position and size of the target in the subsequent video frames. Since the imaging method of infrared images does not depend on the...

AI Technical Summary

Problems solved by technology

[0004] However, most existing trackers only consider the apparent features of the target in the current frame, and use too little inter-frame information, which wastes video feature information and also reduces the performance of the tracker.

Mobile robots often have scenarios where they track fast-moving objects, and these trackers perform poorly in such scenarios

Although some trackers use optical flow to improve performance, the optical flow characteristics they use are off-the-shelf and have not been trained for tracking problems, so the optical flow information is not fully utilized

Images