摘要
针对火灾视频图像中一般亮点区域较小、背景区域图像较大的特点,采用区域分割算法提取各亮点区域并进行处理,提高了算法的抗干扰性,使相同算法的处理时间平均减少到经典算法的1/4,使复杂算法达到实时性要求。同时,在区域分割算法的基础上,成功提取了疑似区域的边界链码,计算了各区域的圆形度。通过神经网络进行数据融合,进一步降低火灾探测系统的误报率。
According to the background in most detecting image hold the great mass of the image ＇ s area, the algorithm of area segmentation is used to scavenge each enlightened areas, which makes the average processing time of the algorithm lessen to 1/4 comparing the classic algorithm. Based on the algorithm of area segmentation, successfully scavenge the Boundary Chain Code of the enlightened areas, and calculate out the roundness of each enlightened areas, and a new mothod for calculating the fire shape is brought forward. Then, Neural Network is used for data fusion and pattern recognition, which reduces the rate of misinformation more in the fire detection system.
出处
《低压电器》
北大核心
2006年第1期32-35,共4页
Low Voltage Apparatus
关键词
火灾探测
图像处理
区域分割
fire detection
image processing
region segmentation
