摘要
通过分析早期森林火灾烟雾图像的运动特征,提出了一种基于数字图像处理技术的森林火灾火源点确定方法。通过对早期报警图像RGB模型进行分量提取并各自直方图均衡化,获得了其中烟雾区域和背景区域的灰度特征,进而根据烟雾区域的连通性选用区域生长法则,分割出独立的目标图像,最终使用深度优先搜索算法遍历整个烟雾区域,确定其最低像素点,即火源点,完成林火的自动定位。结果表明,此种定位方法的准确率达到90.8%。
This paper is aimed at presenting a new method for timely homing forest fire spots based on the digital image detecting technolo- gy so as to be able to extinguish it at an early stage of the fire emerge and spread. In proceeding with our research work, we have analyzed the motion features of the early forest fire smoke image. As the first step, we have tried to apply the histogram equalization method to componential extraction and each the RGB image model pretreated with the early alarming image. Then we have tried to work out the gray characteristic features of the smoke-emerging region with the im- age background region. According to the smoke region connectivity, we have chosen the regional growth rule in our paper to extract the smoke region samples we need instead of the threshold segmentation. In so doing, it would be possible for the forest runners to segment and then obtain the comprehensive smoke image of the fire. And, finally, it would be possible for us to establish a in-depth search algorithm to traverse the entire smoke area so as to make out the lowest pixel as the source of fire and accomplish the automatic paneling of the forest fire. Combined with the practical positioning problem of the forest fire, the actual pixel distance can be defined in less than 10 pixels （including 10 pixels） as the accurate locating. At this stage, pro- cessing 120 warming pictures in this passage, it would be possible for the forest runners to get a positioriing accuracy at about a rate of 90.8 % . Generally speaking, the number represents a good result. In brief, the method we have developed is just the key series for auto- matic paneling of the forest fire under the automatic identification sys- tem. At the same time, we have provided the pixel coordinates of fire for camera calibration technique of forest fire paneling approach. Thus, the results of the trial application show that the method we have developed can spot the forest fire much more quickly and accu- rately than by using other positioning methods, w
出处
《安全与环境学报》
CAS
CSCD
北大核心
2014年第1期50-55,共6页
Journal of Safety and Environment
基金
国家林业局重点科研项目(2010-01)
北京市教委共建项目
关键词
林业工程
森林防火
数字图像处理
自动定位
forest engineering
forest fire prevention
digital imageprocessing
automatic positioning
