Objective Industrial activities serve as the core engine of economic development, yet they also introduce increasingly severe environmental challenges. Sulfur dioxide (SO₂) emitted from industrial stationary sources represents a significant source of atmospheric pollution. In 2023, SO₂ emissions from industrial exhaust gases in China reached 1.803 million tons, accounting for 75.8% of the nation’s total emissions. This substantial share highlights the crucial importance of accurate and continuous monitoring for effective pollution prevention and control. In response to this need, researchers domestically and internationally have developed various optical remote sensing technologies for SO₂ monitoring. Among these, the SO₂ UV camera has gained rapid adoption due to its high spatial and temporal resolution as well as detection accuracy. It has been extensively applied in monitoring emissions from volcanoes, industrial stacks, and ships. However, this technology relies on scattered sunlight as its signal source, which greatly restricts its accuracy during nighttime or under low-light conditions. This limitation impedes the achievement of continuous, round-the-clock monitoring, thereby creating a critical gap in emission surveillance capabilities.

Methods To address these challenges, a self-developed dual-channel infrared imaging system was designed and implemented. The signal channel is equipped with an 8.6 μm filter specifically selected to capture SO₂ gas emission information, while the reference channel employs a 10.2 μm filter to eliminate interfering radiation effects from other substances. To enhance the contrast between the target plume and the background, all original images undergo a comprehensive preprocessing procedure. Both atmospheric background radiation and the intrinsic infrared radiation of the instrument itself were identified as factors that could significantly compromise monitoring accuracy. To mitigate these effects, a background reconstruction method was applied to effectively separate and remove interference stemming from the atmospheric environment and the equipment’s own radiation. This process allows for the extraction of a more accurate characteristic radiation value specific to the SO₂ plume. Subsequently, the SO₂ characteristic radiation signal was isolated through a precise two-channel differential processing technique. This extracted signal was then used to invert a detailed two-dimensional distribution image of SO₂ column concentration. For quantifying the SO₂ emission flux, an optical flow algorithm—a technique derived from machine vision—was utilized to calculate the movement velocity of the plume, enabling a robust estimation of the emission rate. Finally, to conduct a comprehensive evaluation of the system’s overall detection capability, the Minimum Resolvable Gas Concentration (MRGC) model was introduced and applied to analyze system performance under varying conditions.

Results and Discussions Experimental data were collected at a refinery located in Yantai on December 16, 2024, and subsequently subjected to detailed analysis. The results demonstrate that the system successfully captured a clear infrared plume image emanating from the target industrial stack (Fig.4). Furthermore, it effectively generated an inverted image mapping the spatial distribution of SO₂ concentration downwind of the stack (Fig.6). Based on this concentration imagery, the SO₂ emission rate was quantitatively derived (Fig.7). The calculated emission rate exhibited dynamic temporal fluctuations, varying between 10 g/s and 30 g/s throughout the observation period, with a determined average value of 20.1 g/s. Analysis performed using the MRGC model revealed that the integrated detection capability of the system increased significantly in correlation with rising flue gas temperature. This trend aligns well with the inherent attributes of high-temperature exhaust typical of industrial sources. Concurrently, the analysis also indicated that the system’s detection capability could be substantially improved by reducing the monitoring distance between the monitoring system and the emission source (Fig.8), providing a practical operational guideline for field deployment.

Conclusions The infrared imaging remote sensing monitoring system developed in this study demonstrates strong capabilities for monitoring SO₂ gas emissions from industrial stacks. It achieves high spatial and temporal resolution while offering genuine all-weather detection functionality, effectively overcoming the limitation of light dependence associated with UV cameras. The system enables dynamic, continuous, and quantitative monitoring of SO₂ emissions, providing a reliable technological solution for environmental regulation. Its performance underscores significant application value and notable technical advantages within the field of industrial emission monitoring, indicating strong potential for broader implementation in support of air quality management and pollution control objectives.