Accurate 3D terrain modeling by range data fusion from two heterogeneous range scanners

In this paper, we propose a new method for range data fusion from two heterogeneous range acquisition systems (i.e. Laser range scanner and Microsoft Kinect) for the extraction of accurate, realistic and rapidly 3D terrain surface. First we present an active contour model for unsupervised segmentation of RGB image and corresponding depth image by minimizing the total energy function. An unsupervised method for effective curve initialization is based on an equally divided n-regions mean and variance of saliency map. We have transformed both range sensors 3D data to transformed reference frame using PCA algorithm and apply the ICP algorithm to align both data in the reference frame. The alignment of both range scanner data in the transform reference frame is 100 times faster than directly apply the ICP algorithm. To generate the highly precise 3D fused data, we have selected the coarser detailed region of terrain from Kinect and fine detailed region from the Laser range scanner. The 3D fused terrain surface is generated using a Delaunay triangulation algorithm. Finally, the experimental results demonstrate the highly robust and precision of the proposed approach. Therefore, we have generated a seamless integration of terrain 3D surface from two different range scanners 3D range data which validate the correctness of the real 3D surface model of the terrain.