Integration of multiple registered range images finds applications in numerous areas. While most existing integration algorithms detect corresponding points between neighbouring views based on interpoint distances, we define in this paper corresponding area for each point and then carry out geometrical, rigidity, and orientation tests for the same purpose, not only reflecting the scanning nature of sampling resolution, but also the geometry of object surface. To more accurately fuse corresponding points, while most existing algorithms consider the confidence of points, we consider an accumulative fusion scheme of confidences, reflecting the number of times of observation. The fused points are finally triangulated using the improved Delaunay triangulation method, guaranteeing a watertight surface. A comparative study based on real images shows that the proposed algorithm is accurate and efficient and is able to compensate both the accumulative registration errors and imaging noise.