Abstract: We propose a WaveOptima algorithm based on an improved WaveFront algorithm to address issues such as excessive path turns, high overlap coverage, and low planning efficiency in 3D underwater dam detection. The algorithm uses voxel surface adjacency line-of-sight projection technology to extract surface viewpoints and optimize viewpoint layout, reducing approximately 40% of redundant viewpoints. We adopt a 26-neighborhood search strategy to reduce the number and length of path points, in order to improve path smoothness and planning efficiency. We introduce directional and planar costs into the cost function to guide the underwater robot's movement direction and reduce energy consumption. A pre-set direction vector avoids local optima, and a dynamic dead zone escape strategy combined with a motion priority strategy resolves dead zone issues in complex structures. Experimental results show that compared to the traditional WaveFront algorithm and the improved WaveFront algorithm, the WaveOptima algorithm reduces path quantity, overlap coverage, and planning time by 40%, 15%, and more than 38%, respectively, while significantly improving path smoothness and planning efficiency. It is suitable for various dam structures.