Fire flame detection based on feature enhancement and multi-level fusion
In order to improve the recognition and detection performance of fire flame methods,a lightweight fire flame detec-tion model based on feature enhancement and multi-level fusion was proposed by combining traditional image processing with neural network.Multiple color space conversion algorithms were used in the model to enhance the flame feature information,and a two-stage multi-level feature extraction fusion structure was designed,which was combined with spatial attention mecha-nism to extract the flame information from rough to fine.At the same time,based on the characteristics of fire flame,an adap-tive multi-scale fusion structure that gradually integrated from shallow to deep was introduced to improve the detection accura-cy of fire objects in different stages.The results show that the proposed model can effectively improve the detection effect of fire flame,and has higher stability and robustness,which can accurately and efficiently achieve the fire flame detection.The research results can provide more accurate identification results for existing fire detection equipment,so as to better prevent fire accidents.
fire flame detectionneural networkfeature enhancementmulti-level fusionadaptive multi-scale
NETL
NSTL
万方数据
维普
