Target tracking method and system based on fully convolutional Siamese network based on multi-layer feature fusion

Abstract

The invention discloses a target tracking method and system based on a convolution twin network based on multi-layer feature fusion. The method includes: cutting out target template images and search area images of all images in the image sequence training set according to the target position and size of the image , the image pair composed of the target template image and the search area image constitutes a training data set; construct a convolutional Siamese network based on multi-layer feature fusion; based on the training data set, train the convolutional Siamese network based on multi-layer feature fusion to obtain training Good convolutional Siamese network based on multi-layer feature fusion; use the trained convolutional Siamese network based on multi-layer feature fusion for object tracking. In the process of tracking the target, the present invention fuses score maps of different layers, combines high-level semantic features and low-level detail features, better distinguishes interference from similar or similar targets, and prevents target drift and target loss in the tracking process.

Description

Technical field [0001] The invention belongs to the cross field of digital image processing, deep learning and pattern recognition, and more specifically, relates to a target tracking method and system based on a convolutional twin network of multi-layer feature fusion. Background technique [0002] Target tracking has a very important position in computer vision. However, due to the complexity of natural scenes, the sensitivity of targets to changes in illumination, the requirements for real-time and robust tracking, and the existence of factors such as occlusion, posture and scale changes , Making it difficult to track down the problem. Traditional target tracking methods cannot extract rich features from the target so that the target is strictly distinguished from the background, and tracking drift is prone to occur, so the target cannot be tracked for a long time. With the rise of deep learning, general convolutional neural networks can effectively extract rich features of t...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the defects of the prior art, the purpose of the present invention is to solve the technical problems of tracking drift and target loss caused by similar background interference in the prior art

Images