Tomographic synthetic aperture radar (TomoSAR) is used to achieve three-dimensional imaging. Existing methods mainly focus on super-resolution without ambiguity. However, the baseline is uncontrollable in ground-based lunar mapping, which leads to long sparse non-uniform baseline with altitude ambiguity. To suppress the altitude ambiguity, this paper proposes an innovative dual-frequency de-ambiguity method. The mechanism is to enhance real targets and suppress the spurious targets by multiplying the images at different frequency points. Specifically, genetic algorithm (GA) is applied to achieve the optimum frequencies with the lowest peak sidelobe ratio. The computer simulation and real data experiments verify the effectiveness of the proposed approach.