Automated Removal of Planar Clutter from 3D Point Clouds for Improving Industrial Object Recognition

Abstract

The industrial construction industry makes use of prefabrication, preassembly, modularization and offsite fabrication (PPMOF) for project execution because they offer a superior level of control as compared to on-site operations. This control is enabled by systematic and thorough performance feedback loops. Improvement of the feedback systems within these facilities will require a transition away from suboptimal manual data collection to more reliable automated data collection and processing. Laser scanners are an effective tool for automatically gathering dimensional data but extraction of useful information from point clouds remains a challenge. The speed of 3D object recognition methods depends on the size of the search space. Methods for reducing this search space are needed in order to improve the performance of 3D object recognition and subsequent information extraction. Large planar objects (e.g. floors and walls) constitute a large portion of the search space in fabrication facilities, yet are rarely the objects of interest for analysis. In this paper, an automated framework for detecting and removing large planes in point clouds is presented to speed up object recognition. The raw point cloud is first Guassian mapped to normal vector space by calculating normal vectors at each point. The Gaussian sphere is clustered using a density-based clustering algorithm and major parallel planes are segmented from the rest of the point cloud. The major planes are removed and the remaining objects in the scene continue on to 3D object recognition. Results show the algorithm for automatic plane removal can reduce the search space for object recognition by as much as 60% or 70%.

Automatic Super-Surface Removal in Complex 3D Indoor Environments Using Iterative Region-Based RANSAC

“…However, with RGB-D data, parametrically modeled objects (e.g., planes, spheres, cones, cylinders, and cubes) are far more reliably detectable. As a result, researchers have attempted to segment or remove large planar surfaces (e.g., walls, ceiling, and floor surfaces) as a preprocessing or fundamental step before all other algorithms (e.g., [ 27 , 28 , 29 ]).…”

“…Despite the efficiency of clustering feature-based methods, employing multi-dimensional features in large point clouds is computationally intensive [ 28 ]. Furthermore, they are sensitive to noise and outliers [ 43 ].…”

Fast Planar Detection System Using a GPU-Based 3D Hough Transform for LiDAR Point Clouds

“…However, the pre-processes require multiple iterations before executing plane detection so that the time efficiency is affected especially in initial phase [7]. Therefore, the preprocesses of density balancing and re-order registration methods are not adopted in real-time plane surface detection applications [8].…”

Real-Time Plane Detection with Consistency from Point Cloud Sequences

“…Thus, planar structure recognition, which can be formulated as the plane detection problem, has become an important research topic in computer vision for decades. The detected planes, which can be regarded as the abstracted form of an actual scene, contain a lot of high-level structure information and they can benefit many other semantic analysis tasks, like object detection [ 1 ], self-navigation [ 2 ], scene segmentation [ 3 ], SLAM [ 4 , 5 ], robot self-localization [ 6 , 7 , 8 ], For instance, the robot can better map the current environment with the plane detection result, which significantly reduces the uncertainty in the mapping results and improves the accuracy of positioning.…”