一种实时火灾检测算法研究及其嵌入式应用

Real-time outdoor fire detection algorithm and its embedded application

针对户外场景中火灾难于发现的问题,提出了一种基于图像处理的实时火灾哨兵金字塔检测算法。该算法采用哨兵分级策略,在保证报警准确度的同时,较常见的基于逐像素背景差和帧差法视频检测算法运算效率提高数十倍,适用于较大范围内的实时火灾检测,具有良好的应用前景。该算法被成功应用到Hi3516C视频SoC芯片中,成为了某地区秸秆焚烧巡查的重要监测手段,实际报警准确率达94.6%。

A real-time fire detection algorithm based on image processing which is named sentinel pyramid detection algorithm is proposed to help discover fire in the outdoor scenes. The algorithm adopts a sentinel classification strategy to ensure its accuracy and efficiency. It can execute many times faster than other pixel-by-pixel algorithms based on background difference and frame difference. It is suitable and applicable for a wide range of real-time fire detection and has a good prospect. The algorithm has been successfully applied to Hi3516C video SoC chip and has become an important detection way of straw burning because of high accuracy rate of 94.6%.