基于合成图像的Faster R-CNN森林火灾烟雾检测  

FASTER R-CNN FOREST FIRE SMOKE DETECTION BASED ON SYNTHETIC IMAGES

作　　者：张倩 周平平 王公堂[1] 李天平[1] Zhang Qian;Zhou Pingping;Wang Gongtang;Li Tianping(School of Physics and Electronic Science, Shandong Normal University,250358, Jinan, China;Yancheng Biotechnology College,224051,Yancheng,Jiangsu,China)

机构地区：[1]山东师范大学物理与电子科学学院,济南50358 [2]盐城生物工程高等职业技术学校,江苏盐城224051

出　　处：《山东师范大学学报（自然科学版）》2019年第2期180-185,共6页Journal of Shandong Normal University(Natural Science)

基　　金：山东省研究生教育质量提升计划项目(SDYY18064)

摘　　要：本文采用合成图像的FasterR-CNN对森林火灾烟雾进行检测,避免了传统视频烟雾检测方法中复杂的人工特征提取过程.合成烟雾图像是将真实或模拟烟雾插入到森林背景中,解决了训练数据缺乏的问题.将真实合成烟雾和模拟合成烟雾分别训练后的模型放在由真实火焰烟雾图像组成的数据集中测试,测试结果表明,模拟烟雾是更好的选择,模型对薄烟不敏感.通过改进森林火灾烟雾图像的合成过程或者将这个解决方案扩展到视频序列中,可以进一步提高它的性能.In this paper, Faster R-CNN was used to detect forest fire smoke to avoid the complex manually feature extraction process in traditional video smoke detection methods. Synthetic smoke images are produced by inserting real smoke or simulative smoke into forest background to solve the lack of training data. The models trained by the two kinds of synthetic images respectively are tested in dataset consisting of real fire smoke images. The results show that simulative smoke is the better choice and the model is insensitive to thin smoke. It may be possible to further boost the performance by improving the synthetic process of forest fire smoke images or extending this solution to video sequences.

关 键 词：森林火灾烟雾 FasterR-CNN 图像视频烟雾检测 深度学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]