In order to reduce the influence of the joint error of three-dimensional (3D) laser scanner′s multi-sensor point cloud, a simple method to correct the joint error layer by layer is presented, which is based on a columnar standard detected object. The standard detected object is scanned, and the scanning data are fitted to get each section′s coordinate of the circle center. The section′s true value circular function is got by using the actual radius of the cylinder. Translational rigid transformation between each layer′s measured data and true value circular function is done to get this layer′s joint error and corrected value. In order to reduce the influence of random error, repeated measurements are done, and the average corrected value is calculated. Then, some verification experiments are conducted on the measured data of cylinder, cuboid and plaster mannequin, which are placed in different locations. The comparison of stitching result between before and after the correction is made from two aspects of ocular observation and quantified data analysis of sampling sections. The results show that the corrected point cloud can joint together more smoothly, and the relative errors of measured data become lower prominently.