Deep Temporal Network for VOC Gas Recognition in the Middle Infrared Spectrum

The Volatile organic compounds (VOC) gas leak can have harmful effects on both the environment and human health. However, traditional VOC gas leak detection technologies such as electrochemical gas sensors are limited in the outdoor space. We make combination of the advanced Optical Gas Imaging (OGI) system and the deep-learning based temporal convolution network to overcome the drawbacks of the above methods. The Gas Leak Temporal 3D Network (GLT3DNet) are demonstrated in this work, which involves several layers of temporal 3D convolutional operator and identifies the unique spectral signatures of various VOC gases. The synergy of GLT3DNet and the OGI system provides a comprehensive solution for gas leak recognition in infrared videos. The high performance of GLT3DNet enables the OGI system to achieve state-of-the-art results on real-world data captured in various chemical industrial plants.