Single target tracking method based on residual regression network

Abstract

The invention discloses a single target tracking method based on a residual regression network. The method comprises the following steps: 1, preprocessing original training data; 2, inputting the preprocessed data into a residual regression network model, carrying out parameter training, and determining network parameters; 3, after the network parameters are determined, preprocessing the video sequence to be tracked according to the same mode as the step 1; and step 4, inputting a preprocessing result of the video sequence to be tracked into the residual regression network model to obtain a tracking result. The present invention increases speed and allows real-time tracking of objects. Compared with the prior art, the method provided by the invention has the advantages of effectively solving the problems of gradient dispersion and network precision, effectively restraining the problem of precision reduction, reducing the training difficulty of a deep network, greatly improving the precision of single-target tracking, and providing a new direction and idea for solving the problems in the field of subsequent single-target tracking.

Description

technical field [0001] The present invention relates to deep learning and single target tracking, in particular to a single target tracking method based on residual regression network. Background technique [0002] Target tracking technology has flourished since the end of the last century and has become mature and widely used in areas such as unmanned driving, information security, human-computer interaction, and artificial intelligence. In a certain video, with the various changes in the surrounding environment, the process of making the computer automatically recognize and track the target object through a pre-written algorithm is the main task of target tracking. With the rapid development of society, object tracking is more and more required by all walks of life, and the prospect of application is getting better and better. It is precisely because of this that object tracking has become popular in the field of computer vision and has become mainstream. Target tracking...

AI Technical Summary

Problems solved by technology

If only the position and size information of the target object in the first frame of the picture in the picture stream is used as training data, the training effect is not good enough, and the accuracy is far from the result we want.

The existing target tracking algorithms based on deep learning propose methods to solve the above problems from many different perspectives, but there is still a lot of room for improvement in the speed and accuracy of target tracking.

Images