Ice navigation is currently one of the important issues in path planning studies. Most studies are using sea ice models or ice charts for large-scale path planning, but the weather and ice conditions in the polar waters usually change quickly and are not easy to predict. Under these conditions, small-scale path planning that can be adjusted in time for sea ice collision avoidance is needed. This study uses marine radar imaging to reconstruct the ice navigation scene, and establishes a visual-ization function of sea ice warning on the radar image. We combine the bidirectional rapidly-exploring random tree (B-RRT) algorithm and the greedy algorithm to propose a path planning algorithm that generates various path planning schemes for ice navigation. After a number of path planning schemes are quickly generated, the path risk index proposed in this study can be used to evaluate and choose a relatively optimal path planning scheme with lower risk, as well as serve as an assistant reference for path planning decisions in ice navigation.