Multi-Modal Imaging System for Real-Time Liquid Leak Detection

油气管道完整性管理计划(IMP)是为了确保能源运输基础设施的安全运行。及早发现地面设施的微小泄漏是IMP的一个关键要素，对环境安全和保护至关重要。挑战来自动态环境中复杂的液体泄漏情况。一种能够准确、及时地检测到这些泄漏的智能泄漏检测系统有望解决这个问题。 这项研究旨在开发计算算法，以提高基于机器视觉的泄漏检测的准确率和降低误警率。这一由可见光和红外成像组成的智能视觉系统将采用先进的深度学习技术和工业级硬件，以实现对管道基础设施的持续可靠监控。持续监测可以更快地检测泄漏，减少泄漏对环境的影响，并提高高风险地区的管道性能。因此，这项研究将有助于加拿大能源部门的健康增长，为全国各地的加拿大人提供无数的经济利益。

The oil and gas pipeline integrity management program (IMP) is to ensure the safe operation of the energy transportation infrastructure. The early detection of a small leak from above-ground facilities is a key element of the IMP and paramount for environmental safety and protection. The challenges come from the complicated situations of liquid leaks in a dynamic environment. An intelligent leak detection system, which is capable of detecting those leaks accurately and timely, is expected to address this issue. This research aims to develop computational algorithms, which can enhance the machine vision based leak detection with higher accuracy and lower false alarm rate. This intelligent vision system, which is comprised of visible and infrared imaging, will employ the advanced deep learning techniques and industry-level hardware to enable continuous and reliable monitoring of the pipeline infrastructure. Continuous monitoring allows detecting leaks quicker, reducing the environmental impacts from leaks and improve pipeline performance in high-risk areas. Thus, this research will contribute to the healthy growth of Canada's energy sector that provides countless economic benefits to Canadians across the country.