To address the limitations of traditional occluded target imaging technologies，such as low resolution and poor recognition performance，a LiDAR-based 3D point cloud reconstruction technique for occluded targets is proposed in this paper. The method collects point cloud data from multiple perspectives to acquire precise datasets of the target hidden behind the occluder. First，an adaptive density-based clustering algorithm is adopted to make point cloud segments，refining subsets by filtering out irrelevant data to yield potential target fragments. Then，an improved ICP algorithm，combining ISS feature points with a KD-Tree，aligns and merges point clouds from multiple views. Subsequently，with a vehicle as the target，experiments are conducted with 70% coverage using a camouflage net. The results demonstrate that the algorithm effectively extracts and reconstructs the target's contours，achieving an extraction accuracy exceeding 96% and a dimensional error of less than 5%.