Research on infrared image gas leakage detection method based on spatiotemporal multifeature fusion

To address the issue of false alarms potentially caused by environmental interference in infrared gas leakage detection, this paper proposes a detection method based on spatiotemporal multi-feature fusion. By jointly analyzing spatial static features (such as shape) and temporal dynamic features (such as area variation and motion trajectory) of the gas, the proposed approach significantly enhances detection performance. The algorithm first preprocesses the infrared image sequence to improve target visibility, then employs the Visual Background Extractor (ViBe) algorithm to identify motion regions. Finally, a discriminative model is constructed through multi-feature fusion to achieve accurate leak localization. Experimental results show that the proposed method can effectively detect and mark invisible gas leaks while maintaining strong robustness against typical interferences such as pedestrians. This method provides a reliable solution for industrial safety monitoring applications.