Bollard Segmentation and Position Estimation From Lidar Point Cloud for Autonomous Mooring

Abstract

This paper presents a computer-aided object detection and localization method from lidar 3D point cloud data. This topic of interest is in the framework of autonomous mooring, where the ship is tied to the rigid structure onshore (bollard) for autonomous maritime navigation. Using shape and features priors, unlike matching the whole object template to the experimental 3D point cloud representation of scene, two customized algorithms -(a) 3D feature matching (3DFM) and (b) Mixed feature-correspondence matching (MFCM) are presented. The proposed algorithms discriminate and extract the 3D points corresponding to the non-cooperative bollard's surface from background thus capable of classification, localization, and representing it using a unique co-ordinate in the 3D world. The proposed algorithms are tested and validated by implementing upon an experimental data set of 105 scenes where the bollard is at different positions and orientations with respect to lidar mounted on robotic arm. Statistical and probabilistic based approaches are taken into account to determine the performance of proposed algorithms. Model parameters estimation imply that errors resulting from 3DFM algorithm follow homoscedastic bimodal Gaussian distribution with individual Gaussian components having mean 0.03 m, and 0.09 m and both have equal standard deviation of 0.01 m. Further, posteriori component assignment probability is used to identify and cluster the scenes that contribute to relatively larger errors. Finally, an improved algorithm, MFCM is proposed whose errors follow unimodal Gaussian distribution with mean and standard deviation of 0.03 m and 0.01 m respectively thus mitigating the shortcomings of the former.

Edge-Distributed Fusion of Camera-Lidar for Robust Moving Object Localization

“…To compare our proposed method, we further implemented an early fusion algorithm that is widely used in the literature as reference. Fusing data from LiDAR and camera before performing any other downstream tasks is a very traditional method present not only for object localization, but in several other domains [13], [22], [23]. It overlays each 3D point from LiDAR complete point cloud in the image pixel space, so each 3D point is mapped to its corresponding image pixel.…”

“…Not so commonly, some attempts to estimate 3D object localization from 2D images can be found in the literature [12], but those are highly dependent on the training data distribution. To overcome the limitations, LiDAR sensors offer a dense point cloud of the environment and is considered a complement to vision sensors [13]. While cameras give a rich visual representation in 2D, point clouds give the location of objects in 3D as well as more information about shape and occupancy.…”

Application of Laser Systems for Detection and Ranging in the Modern Road Transportation and Maritime Sector

“…The performance of the proposed system was analyzed during several berthing maneuvers and compared to the performance of the commonly used GNSS-based navigational aid system, proving its usefulness in safe berthing experimentally. Additionally, bollard segmentation and position estimation from lidar point cloud data for autonomous mooring was proposed in [80]. Moreover, the authors in [81] developed a system based on the dual-channel lidar for rotorcraft searching, positioning, tracking, and landing on a ship at sea.…”

Semantic Segmentation for Multi-Contour Estimation in Maritime Scenes

“…While many other sensors, e.g., LiDAR [18], RADAR [27] and SONAR [16], have been proposed, cameras serve as a low-cost and lightweight alternative. Cameras capture the surrounding environment and provide rich features of a scene.…”