ABSTRACT
In recent years, the melting ice flood situation of the Yellow River was grim due to the influence of extreme weather. Therefore, ice flood prevention work become very important. Flood control departments could obtain real-time video streams about the Yellow River’s ice through remote video surveillance systems. On this basis, by using target tracking technology and subsequent video measurement technology, they could get the parameter of the ice velocity, which is needed by the ice flood hazard prediction model for early warning. However, this paper found that there would be a phenomenon of false targets and loss targets in the practical application by using the traditional methods of target detection and tracking. Also it would directly impact on the calculation precision of the ice velocity. The reasons for this problem are the characteristics of the Yellow River ice images as well as the problems of lost frames in the process of remote wireless video streaming transmission. This article considers that introducing the pyramid structure of the L-K optical flow method, which is based on a strong angular point feature point set, can solve this problem. This article further verifies the effectiveness and robustness of the algorithm through experiments.
