红外图像帧序列气体泄漏自动检测与定位研究

With the rapid development of industrial automation and progress of science and technology, the requirements of leak detection on industrial equipment are more and more important. There are some problems on the existing equipment of leak detection and location, such as too low detection efficiency and relying on imports. There are some advantages on detection by using infrared detection technology such as quick response, non-contact measurement and wide temperature rage. Therefore, it will be obtained better detection results by designing a reasonable method on leak detecting and location via infrared image technology. This project studies the method of leak detection and location based on infrared image technology. Firstly, a kind of gas leak heat transfer model is set up during the charging and discharging process by analyzing the temperature field of infrared image. Secondly, on the basis of systematic research on adaptive gas leak detection and location technology for a single frame infrared image, a gas leak detection method is presented via pipeline filter approach with analysis and study on the infrared image sequence. Lastly, an experimental platform is set up with infrared image acquisition and processing, and the developed techniques in this project will be verified by this experimental platform. The research results are expected to offer solid theoretical foundation and technical basis for the engineering practice of gas leak detection technology based on infrared image sequence, and further promote the application process of infrared image technology on leak detection for containers.

随着科学技术的进步和工业自动化的迅猛发展，工业设备对泄漏检测的要求越来越高。国内的泄漏检测定位设备存在着一些亟待解决的问题，如检测速度慢、效率低、依靠进口等。红外检测技术作为一种新型的检测和诊断技术，具有速度快、非接触、测温范围宽等优点。根据气体泄漏温度场的特点，以红外检测技术为手段设计合理的气体泄漏检测定位方法，有望获得更好的检测效果。本项目紧紧围绕红外图像帧序列气体泄漏检测定位方法开展研究。首先，通过对红外图像温度场的分析，建立一类气体泄漏孔传热模型。其次，系统深入地研究单帧红外图像的自适应气体泄漏检测定位技术，并通过对红外图像帧序列的研究，提出气体泄漏检测管道滤波方法。最后，搭建具有红外图像采集与分析功能的气体泄漏自动检测定位实验平台，并验证本项目提出的检测定位方法的可行性。研究成果有望为气体泄漏检测技术的工程实践奠定理论基础和技术支持，推进红外检测技术在泄漏检测中的应用化进程。

随着工业生产对泄漏检测要求的提高，传统的气密性检测已不能满足需求，考虑到红外检测技术响应速度快、检测面积大等优点，基于红外图像的气体泄漏自动检测与定位具有很高的使用价值。本项目针对红外图像帧序列气体泄漏检测定位方法问题展开研究。首先，建立了充放气过程中基于红外图像的气体泄漏孔传热模型；其次，研究了单帧红外图像的自适应气体泄漏检测定位技术，提出多种滤波方法实现单帧红外图像泄漏点定位；进一步，提出了基于红外图像帧序列的气体泄漏检测管道滤波方法；最后，完善了基于红外图像的气体泄漏自动检测定位的实验平台，以验证本项目提出的检测定位方法的可行性。本项目提出了一种自适应核回归方法预测红外图像背景信息，通过差分技术和滤波算法改善泄漏点的检测效果。针对传统泄漏检测方法效率低、鲁棒性差的问题，提出了一种基于改进双边滤波的气体泄漏检测方法。考虑到机器学习中SVM算法完善，将其引入气体泄漏点检测算法，用于实现对目标和背景的拟合，进而实现分离。通过对小波理论的研究，提出了一种基于小波变换的红外图像气体泄漏点定位检测算法，以实现实时性要求。为降低气体泄漏检测定位系统的虚警率和漏检率，提出基于红外图像帧序列的气体泄漏检测方法。研究了差分核回归背景预测的泄漏检测方法和基于分层表决的管道滤波方法，并提出一种基于卡尔曼滤波的帧序列泄漏点跟踪算法。此外，针对红外热像仪成像视角的局限性问题，提出了将图像拼接应用到红外图像的气体泄漏检测系统中的解决方案。