Forest fire monitoring based on machine vision has gradually become an important development direction in the field of forest fire monitoring. Smoke is an important indicator of forest fire monitoring. However, many disturbances such as clouds and similar smoke greatly reduce the accuracy of fire recognition. In response to this problem, this paper proposes a machine vision based forest fire monitoring method that combines dehazed and smoke detection. Firstly, several frame images in the appro- priate sample video are extracted as sample images, and the sample images are dehazed by the Haze-Line based dehazing algorithm. Secondly, the smoke detection method based on the Horn-Schunck optical flow method is used to detect the smoke. The maximum inter-class variance method is used to remove the influence of the pixel quality difference between two adjacent frames on the smoke detection. Finally, diffusion analysis is used to do fire monitoring. Results of simulation experiments and comparative analysis show that the proposed method can detect the trend that smoke area gradually increases with time, so as to monitor forest fire under foggy conditions effectively with higher accuracy and robustness.