This study makes the 3D reconstruction of complex-shaped particles based on the multi-view of 2D images. The complex-shaped calcareous sand particles are taken as samples, and typical complex-shaped particles such as block, strip, and dendrite are reconstructed. The final reconstruction accuracy is characterized by convexity, circularity, aspect ratio, and other shape indicators. In the reconstruction process, the initial projection plane of the particles is rotated around the axis to obtain a series of 2D projection images of the particles, and the boundary coordinates are extracted. Then, the 3D point cloud is used to match the coordinates of the obtained 2D image contours, and the point sets located outside the 2D image contours is deleted so that all points are located within the obtained series of 2D image contours. The point cloud is further materialized and constructed to obtain the 3D solid on this basis. By reconstructing 150 particles of different shapes, it is found that the reconstruction error of more than 90% of the particles is within 10%, among which, the largest error range is dendritic particles with the maximum error of 10.84%, and the minimum error is less than 1%. The method is simple and can effectively construct 3D models of complex-shaped particles with high accuracy.