2D fully convolutional network has been recently successfully applied to the object detection problem on images. In this paper, we extend the fully convolutional network based detection techniques to 3D and apply it to point cloud data. The proposed approach is verified on the task of vehicle detection from lidar point cloud for autonomous driving. Experiments on the KITTI dataset shows significant performance improvement over the previous point cloud based detection approaches.

3D fully convolutional network for vehicle detection in point cloud

This paper introduces a novel, region-growing algorithm for the fast surface patch segmentation of three-dimensional point clouds of urban environments. The proposed algorithm is composed of two stages based on a coarse-to-fine concept. First, a region-growing step is performed on an octree-based voxelized representation of the input point cloud to extract major (coarse) segments. The output is then passed through a refinement process. As part of this, there are two competing factors related to voxel size selection. To balance the constraints, an adaptive octree is created in two stages. Empirical studies on real terrestrial and airborne laser scanning data for complex buildings and an urban setting show the proposed approach to be at least an order of magnitude faster when compared to a conventional region growing method and able to incorporate semantic-based feature criteria, while achieving precision, recall, and fitness scores of at least 75% and as much as 95%.

Octree-based region growing for point cloud segmentation

Earlier work demonstrates the promise of deep-learning-based approaches for point cloud segmentation; however, these approaches need to be improved to be practically useful. To this end, we introduce a new model SqueezeSegV2. With an improved model structure, SqueezeSetV2 is more robust against dropout noises in LiDAR point cloud and therefore achieves significant accuracy improvement. Training models for point cloud segmentation requires large amounts of labeled data, which is expensive to obtain. To sidestep the cost of data collection and annotation, simulators such as GTA-V can be used to create unlimited amounts of labeled, synthetic data. However, due to domain shift, models trained on synthetic data often do not generalize well to the real world. Existing domain-adaptation methods mainly focus on images and most of them cannot be directly applied to point clouds. We address this problem with a domain-adaptation training pipeline consisting of three major components: 1) learned intensity rendering, 2) geodesic correlation alignment, and 3) progressive domain calibration. When trained on real data, our new model exhibits segmentation accuracy improvements of 6.0-8.6% over the original SqueezeSeg. When training our new model on synthetic data using the proposed domain adaptation pipeline, we nearly double test accuracy on real-world data, from 29.0% to 57.4%. Our source code and synthetic dataset are open sourced11https://github.com/xuanyuzhou98/SqueezeSegV2

SqueezeSegV2: Improved Model Structure and Unsupervised Domain Adaptation for Road-Object Segmentation from a LiDAR Point Cloud

Place recognition in 3D data is a challenging task that has been commonly approached by adapting image-based solutions. Methods based on local features suffer from ambiguity and from robustness to environment changes while methods based on global features are viewpoint dependent. We propose SegMatch, a reliable place recognition algorithm based on the matching of 3D segments. Segments provide a good compromise between local and global descriptions, incorporating their strengths while reducing their individual drawbacks. SegMatch does not rely on assumptions of ‘perfect segmentation’, or on the existence of ‘objects’ in the environment, which allows for reliable execution on large scale, unstructured environments. We quantitatively demonstrate that SegMatch can achieve accurate localization at a frequency of 1Hz on the largest sequence of the KITTI odometry dataset. We furthermore show how this algorithm can reliably detect and close loops in real-time, during online operation. In addition, the source code for the SegMatch algorithm is made publicly available1.

SegMatch: Segment based place recognition in 3D point clouds

Convolutional network techniques have recently achieved great success in vision based detection tasks. This paper introduces the recent development of our research on transplanting the fully convolutional network technique to the detection tasks on 3D range scan data. Specifically, the scenario is set as the vehicle detection task from the range data of Velodyne 64E lidar. We proposes to present the data in a 2D point map and use a single 2D end-to-end fully convolutional network to predict the objectness confidence and the bounding boxes simultaneously. By carefully design the bounding box encoding, it is able to predict full 3D bounding boxes even using a 2D convolutional network. Experiments on the KITTI dataset shows the state-ofthe-art performance of the proposed method.

http://www.roboticsproceedings.org/rss12/p42.pdf

Vehicle Detection from 3D Lidar Using Fully Convolutional Network

This paper presents a set of segmentation methods for various types of 3D point clouds. Segmentation of dense 3D data (e.g. Riegl scans) is optimised via a simple yet efficient voxelisation of the space. Prior ground extraction is empirically shown to significantly improve segmentation performance. Segmentation of sparse 3D data (e.g. Velodyne scans) is addressed using ground models of non-constant resolution either providing a continuous probabilistic surface or a terrain mesh built from the structure of a range image, both representations providing close to real-time performance. All the algorithms are tested on several hand labeled data sets using two novel metrics for segmentation evaluation.

A. Frenkel

Papers

On the segmentation of 3D LIDAR point clouds