基于连通区域和SVM特征融合的火灾检测  

Fire Detection Based on Connected Region and SVM Feature Fusion

作　　者：王文豪[1] 陈晓兵[1] 刘金岭[1] 

机构地区：[1]淮阴工学院,江苏淮安223003

出　　处：《计算机仿真》2014年第1期383-387,共5页Computer Simulation

基　　金：国家星火计划资助项目(2011GA690190);江苏省高校自然科学基金项目资助(11KJD520003)

摘　　要：为了尽可能地缩小疑似火焰区域,提高火灾检测的准确性和实时性,提出了把图像运动目标检测应用于火焰检测的问题。首先使用背景减除法提取运动目标,然后使用基于连通区域的面积阈值精确地提取疑似区域和其轮廓,再依据早期火灾的视觉特征,抽取四个特征量,即:相邻帧红色比重平均增长率、面积变化率、形状的平均相似度和圆形度,最后通过SVM融合这些特征量进行综合判别。实验结果表明,上述方法计算速度快,检测效果好,误判率低,具有良好的抗干扰能力,为图像目标优化控制提供了依据。In order to reduce the suspected flame area as much as possible, improve the accuracy and real-time character of the fire detection, this paper proposed the thought that the moving target detection is applied to flame de- tection. First, the movement regions were extracted using background subtraction method. Then the flame＇ s suspec- ted region and its contour were accurately extracted using the area threshold based on connected region. And then ac- cording to the early fire＇ s visual characteristic, four features were extracted, namely： red proportion average rate of increase, area rate of increase, average similar degree of form and the value of round scale of adjacent frames. Final- ly, these features were fused by SVM to detect the fire. The experiment resuhs show the algorithm is of high efficien- cy, good detection effect, low misjudgment rate and good anti-interference capability.

关 键 词：火灾检测 图像分割 轮廓提取 特征提取 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]