To solve the problems of poor centroid recognition and low processing efficiency in 3D reconstruction of parallel chromatic confocal measurement systems， a high efficiency and high precision 3D reconstruction method was proposed in this paper. In this algorithm， target extraction and image splicing were performed on all the experimental images to obtain spliced images. In addition， using the regionprops function in MATLAB， the centroid and its connected area of each measured point were extracted， and the H-value and height relationship was constructed with the color conversion algorithm method. Finally， the interpolation fitting algorithm was used to reconstruct the 3D surface topography of the object. To verify the feasibility of the algorithm， the letter “N” and “E” on a one-yuan coin was measured in the author’s self-built parallel chromatic confocal system. The experimental results show that the axial measuring range of the system is 80 μm. Therefore， the measurement accuracy can reach the micron level， and this algorithm can quickly and effectively reconstruct the 3D surface topography of the sample. Compared with the conventional method， the processing efficiency is 5-6 times higher.