Two-viewing angle ladar data registration is an important content in the research filed of obscure target detection by air to ground. Iterative closest-point (ICP) algorithm provides the theoretical basis for it. But the corresponding points are difficult to be confirmed in the ICP algorithm due to the complexity of ladar data in the obscure condition. Based on the analysis of ICP algorithm fundamental principle and ladar data characteristics in the obscure condition, the application strategies and improvements of the ICP algorithm are proposed in three aspects, which are the support-points selecting, corresponding points matching and the pseudo-points removing. Moreover, a novel pseudo-points removing algorithm is proposed based on a fixed-adaptive overlap rate. Then the concrete steps of two-viewing angle ladar data registration using the improved ICP algorithm are elaborated, and lastly the experimental verification is carried out. The experimental results show that the improved ICP algorithm can effectively register the ladar data in the obscure condition, and has stronger robustness and higher registration accuracy compared with other algorithms.