基于图像识别算法的森林防火系统设计
本文提出了一种针对林区的自然环境下进行火灾识别的具体流程, 并针对识别流程中的每一步骤的算法进行了较为详细的说明, 算法遵循由简到繁的原则, 最大限度地减少了对实时图像的大量运算。识别算法通过差分判断、小波判断、图像提取、特征检测 4 个步骤, 对火焰和烟雾的特性进行了综合考虑, 进而大大降低了误判率。通过对不同自然环境下的多幅火焰和烟雾图片的试验,证明所提出的算法合理可行。但针对林区的大量复杂情况,还需在实践中进一步分析并调整各项参数, 提高稳定性。
参考文献(References)
[1] Rodricks B S. Fire detection systems for the millennium [J]. International Maritime Technology, 2001, 113(1): 24-30.
[2] Beltz R C. Integrating GIS/RS technology in extensive forest inventories in the midsouth US [C]//Wood G, Turner B J ed. Proceedings of Integrating Forest Information over Space and Time. Canberra: Auntech,1992(6): 416-419.
[3] 金华彪, 夏雨人, 张振伟. 数字图像处理在火灾探测技术领域的应用[J]. 微型电脑应用, 2003, 19(5): 25-27.
Jin Huabiao, Xia Yuren, Zhang Zhenwei. Microcomputer Applications,2003: 19(5): 25-27.
[4] 关宇东, 韩 媞, 徐枭宇, 等. 复杂背景下基于分割逼近法的抠像技术研究[J]. 电视技术, 2007, 31(7): 75-76.
Guan Yudong, Han Ti, Xu Xiaoyu. Video Engineering, 2007, 31(7): 75-76."
[5] Dedeoglu Y, T reyin B U, Güdükbay U, et al. Real-time fire and flamedetection in video [C]//Proceedings of IEEE ICASSP’05, Ankara, 2005:669-672.
[6] 帅 师, 周 平, 汪亚明, 等. 基于小波的实时烟雾检测[J]. 计算机应用研究, 2007, 24(3): 309-311.
Shuai Shi, Zhou Ping, Wang Yaming, et al. Application Research ofComputers, 2007, 24(3): 309-311.
[7] Horng Wen-Bing, Peng Jian-Wen, Chen Chih-Yuan. A new image-based real -time flame detection method using color analysis [C]//Proceedings of 2005 IEEE International Conference on Networking,Sensing and Control. Tucson: IEEE Press, 2005: 100-105.
[8] 邓 彬, 刘 辉, 连国云, 等. 基于视频的烟雾检测 [J]. 长沙大学学报,2007, 21(5): 87-89.
Deng Bing, Liu Hui, Lian Guoyun, et al. Journal of Changsha University,2007, 21(5): 87-89.
摘要 传统的森林火灾监测技术包括人工护林监测、飞机航测、卫星监测等。本文提出一种基于普通 CCD 摄像头和瞭望塔上现有的短波无线通信设备的森林防火监测系统, 根据实时图像与参考图像的差分及小波分解结果, 当有异常情况发生时, 提取出火焰及烟雾区域, 判断提取区域是否具有烟雾和火焰的动态特征。若判定发生火灾, 即将压缩后的图像通过短波通信设备传回指挥中心。
关键词 火灾识别; 小波分解; 图像提取; 颜色聚类
中图分类号 TP391 文献标识码 A Abstract The traditional fire prevention technologies often involve
manual monitoring, aerial survey and satellite monitoring, but they donot always work satisfactorily. Afire prevention systemis proposed inthis paper for detectingfire in the forest, based on the camera and theexisting short-wave communication equipment on a tower. First, theresultsofthedifferencebetweenthereal-timeimageand the reference
image and its wavelet decomposition are obtained. The fire and smokeregions are identified in case ofabnormal situations. The segmentationmethod isused, based on colorspace,such asRGBand HSI. Then, theregions are decided by dynamic characters of fire and smoke. Theimage is compressed using JPEG2000 compression standard if fire is
found in the region. Finally, the fire image is transmitted tothe centerroom after compression through the short-wave communicationequipment.Keywords firerecognition;waveletdecomposition;imageextraction555