In response to the challenges of weak features, poor visual saliency, and variable morphology of Volatile Organic Compounds (VOCs), a high-precision gas leakage spectral video recognition algorithm based on time-space-frequency joint denoising and multimodal disparity matching model is proposed in this paper. Firstly, the high-precision identification of VOCs is achieved by mining the intrinsic information of high-dimensional time-space-spectrum data, and then the interpretability of traditional methods is organically combined with the powerful representation ability of deep learning through multi-module cascading joint optimization. Finally, by comparing the proposed gas leakage imaging method with international advanced gas monitoring equipment Sencia and Rebellion under the same conditions, it can be seen that the proposed gas leakage imaging method improves the accuracy of methane gas identification by 46.25% for low concentration, and reduces the false alarms to 1/3 of the original one, which verifies the validity and feasibility of the proposed algorithm, providing strong support for monitoring hazardous chemical leakage in the petrochemical industry.