Detection of the River Water Surface Through the Spatio-temporal Image Analysis on a Live Video and the Generation of Simulated Environment

Abstract

In this study, we propose a method of detecting the water surface through spatio-temporal image analysis and verifying it through simulation. The water surface detection algorithm utilizes computer vision to detect the intensity changes in the water surface and non-water surface on a video stream. By calculating the standard deviation of the changes in image intensity, the water surface can be detected since its standard deviation is usually greater than those of the nonwater surfaces (non-water areas) which are usually fixed objects or buildings. The water surface detection algorithm was successfully developed, and it accurately extracts the water surface area. A simulation environment was built to verify the water surface detection algorithm. The real river environment and the motion of the water surface were simulated using the Unreal TM engine by setting various physical conditions. This verified the algorithm for long-term changes that are difficult to observe in a real environment.