The visibility of calculation between patches in complex 3D scene can reflect the occlusion relationship between patches, which is the premise of calculating the radiation effect between patches due to self-emission or light reflection. In the 3D light field of remote sensing scene radiation distribution calculation, full chain optical imaging refinement simulation and other fields have a wide range of applications. At present, there are some problems in the study of the visibility of calculation between patches in complex scenes. Firstly, existing visibility of calculation method based on viewpoints of perspective is easy to miss some visible areas. Secondly, the number of patches in complex remote sensing 3D scene reaches tens of thousands or even millions, the computational complexity of visibility of calculation between patches reaches O(N2), and the existing serial calculation method is time-consuming and difficult to apply. Therefore, this paper carries out the study of the visibility of calculation and efficiency optimization between patches in complex 3D scenes based on super-calculation. A new method for calculating the visibility between patches based on hemispherical uniform random sampling is proposed. The correctness of the algorithm is analyzed theoretically. The visibility of parallel computing model is designed according to the proposed algorithm and experiments is verified on the super computer. The experimental results show that for the 3D application scene of millions of magnitude, compared with the traditional visibility of serial processing strategy, the parallel processing of 512 nodes of the super-calculation platform "Tianhe II" can achieve an acceleration ratio of nearly 500, which improves calculation efficiency significantly.