In this research, a 3D reconstruction of the stereoscopic camera vision for object recognition was implemented using two paths. The first path was four cameras, The distance between each camera and the object was 11 cm. The second path was three datasets captured for the statue and the building with more cameras ranging from 19 to 57 and of the same model used in the first path. The processing stages were first used by the calibration process that determines the internal and external parameters of the camera to convert the 2D points in the image into 3D points. Where 30 2D images of the chessboard and the k value was obtained. Then, the 2D images captured from four cameras are used to extract features. Then, the SGBM algorithm is used to obtain the estimated disparity map. After applying the algorithm, 3D point clouds and Ply file format will be obtained, which can be opened using MeshLab. The results showed that the more cameras there are, the more point clouds there are for the images after the calibration process. This increases the coverage of the surface to be scanned, allowing for more detailed information about the 3D shape.