Abstract: The characteristics of mine flood video images were analyzed. The mine flood identification and trend prediction methods based on video images were proposed, including flood video dynamic identification, region segmentation, area estimation and trend prediction. The results were verified by experiments. The main conclusions are as follows: ① Both threshold pixel grayscale statistical method and pixel grayscale statistical method can monitor and identify floods. The threshold pixel grayscale statistical method not only can suppresses noise below the grayscale threshold and improve the accuracy of recognition, but also can reduce the pixel grayscale statistics, enhance contrast of a particular pixel grayscale range. ② Both the threshold segmentation method and the video differential segmentation method can segment the image of the flood area, the former is better overall and the latter is more detailed.③ The area of the water inrush area can be estimated and the trend can be forecast based on the segmented flood area image.