Aiming at the problem of 3D reconstruction for the irregular and small sized objects, a method of 3D reconstruction was proposed based on monocular vision and rotating line laser. Hardware platform was constructed which mainly contained monocular camera, line laser and rotating table. Camera calibration, light plane calibration, and rotating axis calibration was performed in sequence to obtain camera intrinsic and extrinsic parameters, light plane formulation and rotating axis formulation, so as to build the relationship between pixel coordinate and world coordinate. Images of line laser reflected from rotating object were captured by monocular camera, then pixel coordinates of image were transformed to point cloud through grayscale centroid method after aera extraction, filter and threshold process, the 3D point cloud of object can be obtained by adding rotating offset to each frame of point cloud. Volume of data can be reduced by down-sampling while keeping the feature of the point cloud, statistical analysis and filtering was performed on the discrete points, mixed sampling method was used for smoothing the point cloud, surface normal was estimated through covariance matrix of the nearest neighbor point, and 3D reconstruction was realized by curving point cloud based on Delaunary growing algorithm. Finally, a lens was reconstructed as the test object and the radius dimensions were measured, test results show that the average error of outer profile radius is less than 0.5mm, the maximum error is less than 1mm, and the standard deviation is less than 0.7mm, which proves that the 3D reconstruction method is feasible and generic under certain demand of precision.