Open AccessTarget tracking based on neural network depth feature and texture fusion
Author(s) -
Ye Cao,
Meiju Liu,
Shanykui Yang,
Guodong Yang,
Peng Chen,
Shuyun Zhu,
Ge Zhuang,
Yuwang Liu
Publication year - 2020
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.179
H-Index - 26
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/461/1/012021
Subject(s) - artificial intelligence , computer science , convolutional neural network , pattern recognition (psychology) , fusion , feature (linguistics) , computer vision , artificial neural network , convolution (computer science) , texture (cosmology) , tracking (education) , set (abstract data type) , feature extraction , image (mathematics) , psychology , pedagogy , philosophy , linguistics , programming language
This paper presents a method of target tracking based on convolution neural network and texture feature fusion. The lower layer of the convolutional neural network can extract some spatial structure, shape and other features of the target. High-level level can extract relatively abstract semantic information. In this paper, vgg-m convolutional neural network is adopted to realize tracking by adaptive fusion of the extracted depth features of Conv2 and Conv5 with the texture features extracted by two-dimensional Gabor filtering. In this paper, the experimental analysis of this method is carried out on the OTB2013 data set, and the results show that this method can achieve more accurate positioning of the target and has a strong timeliness.