Open AccessResearch on flame recognition technology based on local complex features
Author(s) -
Di Zhu,
Kamarul Arifin Ahmad,
Aolin Chen
Publication year - 2022
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 - 10.1088/1742-6596/2246/1/012074
Subject(s) - artificial intelligence , pattern recognition (psychology) , scale invariant feature transform , computer science , feature extraction , feature (linguistics) , fuse (electrical) , computer vision , filter (signal processing) , noise (video) , image (mathematics) , engineering , philosophy , linguistics , electrical engineering
Traditional flame recognition methods based on image features are difficult to extract flame image features effectively, resulting in low flame recognition accuracy, while such methods mostly perform specific flame recognition for specific scenes, and when the scene, flame color and other features change, it is difficult to perform flame recognition effectively. To address this problem, this paper proposes a flame recognition scheme based on local complex features. Its main purpose is to fuse multi-scene flame data, introduce the characteristics of flame in color space through the process of extracting feature descriptors in SIFT, so as to filter the extracted feature descriptors with noise interferers, and transform the extracted feature descriptors into feature vectors by using the key point bag-of-words method, and finally a general fast flame recognition model based on the limit learning machine. In this paper, we explore the upper limit of the capability of traditional image feature representation to pave the way for the application of deep learning to the flame recognition problem.