Video Image Processing Technology Of Snowfall Surveillance

In most regions of China, especially high-latitude areas, heavy snow arises from low temperature and cold wave, and seriously threats the rail traffic safety and operation efficiency in winter. In situ monitoring of snowing and icing disaster along the railway is an important issue to be solved. Research on this issue is significant.First, the physical, optical and time-domain characteristics of snowflakes or particles in video images are analyzed, to establish the discriminating foreground and background model used for the following video image segmentation.Then, the conventional frame difference and background subtraction methods are modified to segment the snow particles from video images respectively, and their segmentation results are analyzed and compared. Due to the color intensity difference of falling snow in various background, an enhanced frame difference method is proposed to compute the frame difference with continuous multi-frame images from video, to eliminate incomplete object detection and cavities, so as to improve the robustness of the segmentation. K-means clustering method is used in the background subtraction method to obtain precise background model and classify the snow and background through their intensity for the purpose of improving the falling snow segmentation of video images. Comparison analysis is made to the results of the two methods, and find that the frame difference method is better than the background subtraction method in respect of real time segmentation, and is more appropriate for various background, while the background subtraction method is applicable to static background, and gains less noise in the result of segmentation.Finally, the color difference between falling snow and background is very small, making snow segmentation difficult. A region of interest from video image is thus selected to segment the falling snow and analyze the snowfall, which is able to overcome the difficulty. Then, a statistical method is used to get the information of snowfall of a surveillance point.Video surveillance and image analysis technology gains the information of snowfall during snow and ice weather, to control the operation speed of high-speed train and provide the rail administration sound scientific evidence of decision making for disaster prevention and mitigation.