Object classification integrating results of each scan line with low‐resolution LIDAR

Abstract

To recognize objects by using low‐resolution LIDAR for autonomous cars, we proposed a method to calculate the independent results for each scan line before integrating them. In the proposed method, objects can be recognized in one learned model even if the number of scan line irradiated to objects is different. This brings an advantage of saving time for preparing some models and learning data. We tried to classify pedestrians, bicycles, motorbikes, cars, and other objects for evaluating the performance, and obtained an accuracy of 99.00%. In addition, we compared the proposed method with a fully connected neural network method and 2DCNN method, and showed the proposed method is more robust against the missing of scan lines. © 2019 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.

Transformer point net: cost-efficient classification of on-road objects captured by light ranging sensors on low-resolution conditions

“…This approach has been present in models where a significant number of 2D rendering extraction techniques have been developed. 2D point cloud slices [6,7], object views [8], elevation images [9,10] and spherical representations [11,12] are some of these, each with diverse versions. 2D views representations usually omit relevant 3D information, which can give a performance drop.…”