There has been a long history for the ice disaster in the Yellow River, and monitoring methods at present relatively backward. Therefore, an effective real-time monitoring method is especially important. Currently we cannot get some important parameters, for example the ice density, which is needed by the hazard prediction model of early warning accurately and validly. This paper will put forward a new method which is to for access the ice density based on the modified C-V model.By Using the C-V model which is joined with gradient information and accelerating factor, we can identify the target and segment the ice images of Inner Mongolia section of the Yellow River which is collected by the UAV. The Yellow River regime is not only more complicated but also have the characteristics of high sediment content. Several sections of Yellow River's water is muddy. Target is difficult to identify. In addition to this the ice shape is very irregular. Therefor by using of the traditional image edge detection technology and original C-V model can fail to effectively identify ice. Nevertheless, the improved C-V model mentioned above can solve such problems effectively. The experimental results also prove that the algorithm is robust and effective.
