Fire smoke detection based on texture features and optical flow vector of contour

Abstract

A fire smoke detection algorithm based on optical flow method and texture feature is proposed, which can be used for early fire alarm. It establishes the image of Pyramid, and uses LBP and LBPV to extract the static texture features of different levels. Due to the turbulence characteristic of the smoke motion, the smoke direction is consistent, and optical flow vector analysis is used to judge the movement directions of the suspicious area contour, to reduce the computational complexity. Finally, the texture features of smoke image are identified by SVM. Simulation results show that the proposed algorithm can ensure the timeliness and accuracy of the algorithm.