基于支持向量机的泄漏气体云团热成像检测方法

Thermal Imaging Detection Method of Leak Gas Clouds Based on Support Vector Machine

基于热成像的气体泄漏检测技术以其检测效率高、直观可视等优点,已成为石油天然气泄漏检测的重要手段,但常规的气体泄漏热成像检测方法需要检测人员从视频图像中主观地判断泄漏气体痕迹,容易发生漏检、误检。研究了一种基于尺度不变特征变换(SIFT)和支持向量机(SVM)的泄漏气体云团热成像检测算法,采用帧间差分法从红外图像序列中筛选目标区域;分别提取泄漏气体和干扰物的SIFT特征;使用SVM对候选区域进行目标判别,提取泄漏气体云团目标。针对真实复杂场景中包含乙烯、甲烷等的气体泄漏图像和运动人员、漂动树木、野草等干扰图像,建立了1000个典型目标图像数据库,通过图像检测仿真,可得所提算法对距10~150 m处的泄漏气体云团的分类准确率可达92.5%。结果表明,采用该检测方法可自动排除其他运动物体的干扰,有效检测出泄漏气体云团。

Gas leak detection technology based on thermal imaging has become an important means of oil and gas leakage detection because of its high detection efficiency and visibility.The conventional methods need personnel's subjective judge to trace gases from the video,so it is easy to lead miss and false detection.Therefore,this paper studies a thermal imaging detection algorithm of leaking gas clouds based on scale invariant feature transform(SIFT)and support vector machine(SVM),and uses the inter-frame difference method to screen the target region from the infrared image sequence.SIFT features of leaking gas and disturbance were extracted,respectively.SVM is used to identify the target in the candidate region and extract the leaking gas cloud.A database of 1000 typical target images was established for real complex scenes,including ethylene,methane,and other gas leakage images and disturbing images such as moving person,trees,and weeds.Through detection experiment,the classification accuracy of the proposed method for leaking gas clouds at 10--150 m can reach 92.5%.The results show that this detection method can automatically eliminate the interference of other moving objects and effectively detect the leaking gas cloud.