Explore chapters and articles related to this topic
Robot Path and Motion Planning
Published in Jitendra R. Raol, Ajith K. Gopal, Mobile Intelligent Autonomous Systems, 2016
In this roadmap (RM) approach, the free C-space (set of feasible motions/CS) is reduced to or mapped onto a network (NW) of 1D lines, the approach also being called the Retraction, Skeleton, or Highway approach [5]. In this RM case, the search for a solution is limited to the NW, and the PMP problem degenerates to a graph-searching problem. There are four popularly known RMs approaches [5]: (i) Visibility graph (VG), (ii) Voronoi diagram (VD), (iii) Silhouette (S) and (iv) Sub-goal NW (SNW). The VG, as the collection of lines in the free C-space, connects a feature of an object to the feature of another object. These features are vertices of polygonal obstacles; there are the edges in the visibility graph, and the VG is constructed in 2D. In the VD, there is a collection of objects, and the VD is a partition of space into numerous cells. Each of these cells consists of the points generally closer to one particular object (than any other objects). In the approach called Silhouette, one projects an object (which is) in a higher dimensional space to a lower dimensional space and then traces out the boundary curves of the projection, similar to tracing out the silhouette of a person. In the SNW method an explicit representation of the configuration obstacles is not built, but a list of reachable configurations from the starting configuration is maintained. On reaching the goal configuration (i.e., when it is reachable) the PMP problem is solved. This reachability (of one configuration from another) is determined by a simple local MP algorithm. This local operator entails the moving of the robot in a straight line between the configurations.
Modified U-Net based 3D reconstruction model to estimate volume from multi-view images of a solid object
Published in The Imaging Science Journal, 2023
Radhamadhab Dalai, Kishore Kumar Senapati, Nibedita Dalai
After obtaining the depth data, the 3D reconstruction model has been utilized to reconstruct the irregular solid objects for volume estimation. However, the exact 3D object cannot be reconstructed using the silhouette method. The number of voxel counting or mathematical modelling are further used to reconstruct the three-dimensional object. However, mesh based model and pixel cloud model are the important methods for three-dimensional reconstruction. In the proposed model, Dense Point Cloud Generation Algorithm [27] based on Patch-based Multi View Stero (PMVS) is employed for reconstructing the 3D point cloud of object. The 3D reconstruction approach comprised of four steps: PMVS point cloud generation, seed-based point cloud expansion, point cloud optimization and outliners. The region of the point cloud begins from a set of seed points and provides a set of seed points as outcome. The output seed point has been utilized and it makes use of projection rules between seed point and pixels for dividing the generated seed points for extending the denser seed points. Further, a patch-based Multiphoto Geometrically Constrained Matching algorithm (MPGC) has utilized for optimizing the extended seed points to determine better accuracy. Moreover, density constraint has been applied for filtering the outliners.
Generation and quality improvement of 3D models from silhouettes of 2D images
Published in Journal of the Chinese Institute of Engineers, 2018
Watchama Phothong, Tsung-Chien Wu, Chun-Yeh Yu, Jiing-Yih Lai, Douglas W. Wang, Chao-Yaug Liao
The curvature at a vertex on the 3D model is evaluated as follows. All 3D vertices are projected onto a 2D image and the boundary contour of the projected vertices is obtained. This boundary contour is called a projected boundary hereafter. Consider a vertex vi on the 3D model and its projection on the 2D image is vip . When vip is on the projected boundary, its closest point vis on the image silhouette is evaluated. Each image silhouette essentially represents a projected profile of the object on a viewing plane. A curvature at vis along this profile can be estimated by evaluating the angle of two k vectors and , directed from vis to k points behind and k points ahead, respectively (Pan, Meng, and Tu 2003). Therefore, the curvature at vis can be regarded as the curvature of vip on this viewing plane. On the contrary, when vip is not on the projected boundary, its curvature is set to 0. When all images are checked, the maximum curvature can be found and is regarded as the curvature of vi. Edge flip is finally implemented in the remeshing process to adjust the regularity of the meshes locally (Figure 9(c)).