Emission rate is a critical parameter for assessing the environmental impact of ships, as it directly relates to the amount of pollutants released into the atmosphere per unit of time. Monitoring and controlling ship emissions has become a pressing global challenge due to the increasing contribution of maritime activities to air pollution and climate change. However, traditional monitoring techniques, while effective in capturing spatial pollutant concentration distributions, face inherent limitations in their physical principles. These methods rely on localized or single-point measurements, making them unsuitable for accurately capturing dynamic processes like plume transport and dispersion. As a result, they fail to deliver precise real-time emission rate measurements.To address these limitations, this study leverages the unique advantages of UV cameras for two-dimensional pollutant imaging and proposes an innovative method for calculating ship exhaust emission rates using the Farneback optical flow algorithm in machine vision. UV imaging enables spatially resolved detection of pollutant concentrations over a large field of view, offering a significant improvement over traditional point-based methods. The Farneback optical flow algorithm, which has been widely validated for rigid body motion estimation, is adapted to analyze the dynamic behavior of fluid plumes in this study.The SO₂ UV camera utilizing the specific absorption characteristics of SO₂ molecules in the UV spectrum. By recording the spectral signals generated from SO₂ absorption of UV light and converting them into visualized images, the camera provides precise detection and localization of SO₂ emission sources. However, SO₂ molecules exhibit significant absorption of solar radiation in the 280~320 nm wavelength range, where black carbon particles in the exhaust also introduce absorption interference. To address this, a dual-channel imaging framework is proposed. This framework, based on the Beer-Lambert law and the differential absorption cross-sections of SO₂ and black carbon particles, includes both monitoring and correction channels. By calculating the optical thickness difference between these channels, the system effectively eliminates black carbon interference, yielding accurate SO₂ optical thickness measurements. Additionally, a spectrometer channel is incorporated to establish calibration curves, correlating SO₂ optical thickness with gas concentration for precise retrieval of SO₂ concentrations in ship exhaust.Building on these principles, a dual-channel UV remote sensing imaging system was developed to enable quantitative monitoring of ship exhaust emissions. Field experiments were conducted at Yantai Port, targeting a passenger ro-ro ship, to collect image data of exhaust plumes under real-world conditions.The Farneback optical flow algorithm was employed to analyze the SO₂ concentration image sequences. By using consecutive concentration images as input, the algorithm calculated pixel-wise motion vectors based on the assumptions of constant local gradient and smooth optical flow. This analysis produced a dense optical flow field, providing detailed information on the flow direction and velocity of the exhaust plume. To calculate the emission rate, the cross-sectional region of the plume was extracted, and the SO₂ concentration values for each pixel were combined with the velocity vectors from the optical flow field. A discrete convolution process was applied, allowing the pollutant emission rate to be determined over specific time intervals. The integration of high-resolution UV imaging and machine vision-based optical flow analysis provides a robust and scalable framework for real-time emission rate monitoring.The results of this study validate the reliability and effectiveness of the proposed approach for SO₂ concentration measurement and emission rate calculation. The integration of the Farneback optical flow algorithm with UV imaging enables precise, real-time tracking of dynamic emission processes. This method provides critical support for the real-time tracking and quantitative analysis of ship pollutant emissions, enhancing the enforcement of maritime emission regulations. Furthermore, the technical framework and algorithm developed in this study are broadly applicable and can be extended to monitor other atmospheric pollutants, including NO? and particulate matter.