Ship targets are important objects for marine monitoring, and infrared imaging system has been widely used in ship inspection systems due to its feature of working at the same time during the day and night. However, infrared imaging systems will be easily affected by the release of smoke screens, which result in the invalidation of ship detection systems. Therefore, timely and effective detection of the smoke interference area in the infrared ship image is of great significance for accurate ship target detection. Aiming at the problem of detecting the interference of smoke area from ships in infrared images, a smoke detection method based on the fusion of saliency and dense optical flow was proposed in the paper. Because the smoke screen released by the ship was obviously different from the background, the AC algorithm of multi-scale neighborhood filtering was firstly used to detect the saliency area of the image, and the significant smoke screen area was extracted. Then, the movement characteristics of the smoke screen were used to compare the front and back frame of the image sequences, and the frame dense optical flow was calculated to obtain the motion information of the image. By setting the threshold to filter the obvious motion points, expand the motion point area, merge the split motion areas, the motion smoke area was obtained. Finally, the saliency area and the motion area were fused, and the final smoke screen area was obtained. The experimental results show that the method can effectively detect the smoke screen area, and is able to adapt to the changes of the reflected light of smoke screens and the background brightness variations.