A Novel Underwater Single Target Tracking Algorithm based on Siamese Convolutional Neural Network

Abstract

Aiming at the problem of low accuracy of underwater target tracking due to poor contrast and sharpness of underwater visual data and complex background, a novel underwater single target tracking algorithm based on Siamese Convolutional Neural Network was proposed. Through the use of deep convolutional network, the target features could be better extracted to improve the algorithm’s ability to deal with the above problems. First, convolutional neural network is used to extract the features of the template frame image and the detection frame image. Then the convolution features of the two are cross-related by generating the region proposal network. Finally, the classification branch and regression branch are output. The output classification branch is used to classify the background and target, and the output regression branch is used to predict the location of the tracking target boundary box. The experiment shows that the above tracking algorithm can better adapt to underwater environment and improve the accuracy and success rate of underwater target tracking compared with the traditional underwater target tracking algorithm.