Mean Shift: A Robust Approach Toward Feature Space Analysis
IEEE Transactions on Pattern Analysis and Machine Intelligence
Journal of Intelligent and Robotic Systems
Globally Consistent Range Scan Alignment for Environment Mapping
Autonomous Robots
Mean Shift, Mode Seeking, and Clustering
IEEE Transactions on Pattern Analysis and Machine Intelligence
Exploring artificial intelligence in the new millennium
Line Extraction in 2D Range Images for Mobile Robotics
Journal of Intelligent and Robotic Systems
Pattern Recognition
IEEE Transactions on Computers
On measuring the accuracy of SLAM algorithms
Autonomous Robots
Image and Vision Computing
Simultaneous multi-line-segment merging for robot mapping using mean shift clustering
IROS'09 Proceedings of the 2009 IEEE/RSJ international conference on Intelligent robots and systems
Clustering and line detection in laser range measurements
Robotics and Autonomous Systems
A confidence measure for segment based maps
PerMIS '09 Proceedings of the 9th Workshop on Performance Metrics for Intelligent Systems
Prediction-based geometric feature extraction for 2D laser scanner
Robotics and Autonomous Systems
On mean shift-based clustering for circular data
Soft Computing - A Fusion of Foundations, Methodologies and Applications
The estimation of the gradient of a density function, with applications in pattern recognition
IEEE Transactions on Information Theory
Robotics and Autonomous Systems
| Hi-index | 0.00 |
The recent trend of deploying mobile robots in large indoor environments calls for development of efficient map representation techniques. Compared to the more common occupancy grid representation, maps built with line segments are more compact and scale well with the environment size. In this paper, we propose an offline method for building maps of indoor environments using line segments extracted from registered laser range scans. At the core of this method lies a new formulation for identifying and then merging into one, all line segments that represent the same planar surface in the environment. Two successive steps of density-based clustering, applied on the extracted segments, enable us to delineate the segments that are in close proximity to each other and hence represent the same surface. The proposed method has accurately built maps of a wide variety of indoor environments, both real and simulated. Compared to two other similar methods, it has generally produced better maps. We also propose ways by which the goodness of the produced maps, in terms of how closely they resemble the ground truth, can be assessed.