Real-time leak detection using an infrared camera and Faster R-CNN technique

Real-time hydrocarbon leak detection is an essential part of process safety and loss prevention program. Optical gas imaging (OGI) is one of the attractive methods to monitor hydrocarbon leak in the processing system. The manual analysis of a video frame to detect a potential leak is cumbersome and error-prone. The purpose of this study is to develop the automated hydrocarbon leak detection using appropriate technology and numerical technique. This is achieved by integrating Faster Region-Convolutional Neural Network (Faster R-CNN) technique with the OGI technology. The application of the procedure is demonstrated using the videos of the hydrocarbon leaks from an Ethane cracker plant. The videos are used to train the Faster R-CNN and subsequently used for the testing. The performance of the proposed integrated approach is compared with the Single Shot MultiBox Detector (SSD) models. The results confirm the proposed optimal model is superior compared to the SSD models.