Research On The Flame Detection Algorithm Using Image Processing

Fire is a great threat to people’s production and life because of its sudden and uncertainty,especially in the unattended warehouse,exhibition center,workshop and other tall environment.It is necessary to configure fire alarm system.The algorithm of flame detection based on image processing is designed in this thesis,which can recognize the flame through image video and realize the function of fire warning.In this thesis,the overall framework of the flame detection algorithm is given.The rules of flame discrimination based on image pixels are discussed.The algorithm of flame detection based on image processing is designed in detail,which includes algorithms for flame candidate region extraction and flame identification.Extract flame candidate regions by removing flame pixel interference points and merging small areas.The flame identification algorithm based on the deep learning neural network can further identify the flame candidate areas and improve the accuracy of flame detection.The deep learning neural network structure of the flame discriminator is analyzed,and the details of the algorithm implemented by the ARM system are given.Finally,the thesis gives a smoke detection algorithm based on deep learning.It can detect the smoke in the image video,so it can send out fire warning in advance.The algorithm of flame detection based on image processing is tested.The results show that the thesis’ s algorithm has 93.5% accuracy,which meets the needs of practical engineering application.