Open AccessAutomatic keying algorithm for multi-category target recognition
Author(s) -
Liping Mao
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1982/1/012122
Subject(s) - computer science , pattern recognition (psychology) , artificial intelligence , hyperspectral imaging , algorithm , automatic target recognition , feature (linguistics) , sample (material) , fuse (electrical) , artificial neural network , synthetic aperture radar , philosophy , linguistics , chemistry , chromatography , electrical engineering , engineering
In this paper, through an in-depth study of the automatic keying algorithm for target recognition and using multi-class algorithms for its analysis, a saliency detection model based on the hypercomplex Fourier transform is proposed, which can quickly search for information related to the current task requirements. The problem of sample imbalance in deep neural network training exists, the module is used many times to fuse multi-scale features, the loss function uses weighted cross-entropy loss, and the weights are determined according to the proportion of samples in the training sample, which can solve the problem of the model tends to fit the category with more samples. After analysis, the proposed hyperspectral image fast feature enhancement algorithm based on guided filtering can effectively solve the problem of “the same object, different spectrum”, and the classification accuracy of small sample high-dimensional data is improved greatly. At the same time, the complexity of processing high-dimensional data such as hyperspectral remote sensing images is greatly reduced. The experimental results show that the processing time of the proposed fast feature enhancement process for hyperspectral remote sensing images in this paper decreases than that of the direct use of guided filtering.