인공지능 학습을 위한 Multi-Channel LiDAR 성능평가 = Multi-Channel LiDAR Performance Evaluation for AI Training

다국어 초록 (Multilingual Abstract)

The field of artificial intelligence learning is expanding from image-based approaches to 3D lidar sensing for recognition tasks such as mapping, localization, and object detection. This has led to a rapidly growing LiDAR manufacturing industry with several competing manufacturers releasing new sensors on a regular basis. This has resulted in a wide variety of LiDAR with different characteristics such as distance, resolution, field of view, and price, making it necessary to compare the characteristics and performance of each product in more depth. In this study, we established several metrics to evaluate the performance of multi-channel LiDAR commonly used in artificial intelligence training and autonomous driving. We qualitatively identified various open issues with specific LiDARs. Evaluated the accuracy and precision of LiDAR patterns and accumulated point clouds in static and dynamic environments. Considered real-world applications, utilizing pedestrians and walls as additional targets of interest to evaluate clutter that affects AI learning. A thorough evaluation of these lidars was made through their potential application to AI learning.

참고문헌 (Reference)

1 Bresson, G., "Simultaneous localization and mapping : A survey of current trends in autonomous driving" 2 (2): 194-220, 2017

2 Lambert, J., "Performance analysis of 10 models of 3D LiDARs for automated driving" 8 : 131699-131722, 2020

3 Bae, K., "On-site selfcalibration using planar features for terrestrial laser scanners, The International Archives of Photogrammetry" 36 : 14-19, 2007

4 Burgard, W., "Map-based precision vehicle localization in urban environments, In Robotics: Science and Systems III" 121-128, 2008

5 Azim, A., "Detection, classification and tracking of moving objects in a 3D environment" 802-807, 2012

6 Rist, C. B., "Cross-sensor deep domain adaptation for LiDAR detection and segmentation" 1535-1542, 2019

7 Takeuchi, E., "A 3-D scan matching using improved 3-D normal distributions transform for mobile robotic mapping" 3068-3073, 2006