基于改进GMM背景差分方法的蒸汽泄漏检测研究  

Research on Steam Leak Detection Based on Improved GMM Background Subtraction Method

作　　者：鄢家鑫 陈青[1] 李朋洲[1] 周寒 刘晓东 YAN Jiaxin;CHEN Qing;LI Pengzhou;ZHOU Han;LIU Xiaodong(Nuclear Power Institute of China,Chengdu,Sichuan 610213,China)

机构地区：[1]中国核动力研究设计院,四川成都610213

出　　处：《计算技术与自动化》2023年第4期147-153,共7页Computing Technology and Automation

摘　　要：为检测反应堆回路试验装置蒸汽泄漏,提出了一种改进的基于背景差分法的蒸汽泄漏红外视频检测方法。该方法利用红外相机采集图像,使用Wasserstein距离匹配像素点,采用K-means方法去更新改进的混合高斯背景模型(GMM);在后处理中,采用自适应均值滤波及形态学方法抑制环境噪声,通过对泄漏图像灰度形状面积的分层融合特征判断实现对蒸汽泄漏的分级检测。为实现对算法效果的量化,利用泄漏报警率、F1分数等指标评价检测方法的优劣,提出以Dice系数作为算法分割效果的评价函数;在steamTS200inf数据集上的试验结果表明,泄漏报警率可达98.52%,最后一帧分割Dice值可达78.31%,本文方法可有效检测蒸汽泄漏。In order to detect the steam leakage of the reactor loop test device,the paper proposed an improved infrared video detection method of steam leakage based on the background subtraction method.This method collects images with the infrared camera,and uses the Wasserstein distance to match the pixels,and uses the K-means method to update the improved Gaussian mixture background model(GMM).In the post-processing,adaptive mean filtering and morphological methods are used to suppress environmental noise.The hierarchical fusion feature of the grayscale,shape and area of the leakage image is used to achieve graded detection of steam leakage.To quantify the algorithm’s effectiveness,metrics such as leakage alarm rate and F1 score are used to evaluate the detection method,and the Dice coefficient is proposed as the evaluation function for the algorithm’s segmentation effectiveness.Experimental results on the steamTS200inf dataset demonstrate a leakage alarm rate of 98.52%and a final frame segmentation Dice value of 78.31%,indicating the effectiveness of the proposed method in detecting steam leakage.

关 键 词：泄漏检测 背景差分 红外视频 图像分割 

分 类 号：TP29[自动化与计算机技术—检测技术与自动化装置]