摘要
针对目前基于传感器技术的火灾探测器存在检测区域小、干扰大、误报率高等不足,提出了一种基于混合高斯模型与机器视觉相结合的火灾检测方法。该方法通过对摄像头获取的图像信息进行处理和分析,利用火灾早期产生的烟雾和火焰的颜色及形体变化特征来探测火灾。通过建立混合高斯模型对火灾进行识别,再对识别结果进行进一步的动态特征识别。试验证明,该方法克服了传感器检测方法的缺点,实时性好、识别率高。
Aiming at the deficiency of existing fire detectors based on sensor technology,such as small detecting area,large disturbance,and high rate of misreporting,etc.,the new fire detection method based on combination of Gaussian mixture model（GMM） and machine vision is proposed.Through processing and analyzing the image information acquired by camera,the fire is detected by adopting the characteristics of colors of smoke and flame,and the variation of form and structure of the initially generating fire.Then,Gaussian mixture model is built to identify the flame,and the result is further identified dynamically.The test verifies that this method overcomes the defects of sensor detection,and features good real-time performance and high identifying rate.
出处
《自动化仪表》
CAS
北大核心
2012年第3期60-62,共3页
Process Automation Instrumentation
基金
云南省教育厅科学研究基金资助项目（编号：042140D）
关键词
火灾检测
混合高斯模型
机器视觉
自动识别
像素点
Fire detection Gaussian mixture model（GMM） Machine vision Automatic identification Pixels
