@article{ref1,
title="Performance improvement of deep learning object detection method using dynamic occupancy grid map",
journal="Transactions of the Korean Society of Automotive Engineers",
year="2022",
author="Jang, Harin and Cho, Juyeon and Heo, Sejong and Kang, Yeonsik",
volume="30",
number="10",
pages="839-847",
abstract="Dynamic occupancy grid map(DOGM) is a method of representing nearby objects information such as their position and speed on a grid map. This study evaluated the performance of a deep learning-based algorithm developed to detect and classify objects with free probability, velocity, and height information from a DOGM obtained using 3D lidar sensor measurements. The objects are classified into three categories: car, truck and bike. Nuscenes dataset was used to train and evaluate the developed model. The performance of the algorithm is evaluated by calculating the mean average precision(mAP) for the entire three classes, which is compared with other benchmark object detection algorithms such as BirdNet, BirdNet+ and BEV-Net.   키워드: 자율주행, 동적 격자 지도, 객체 인식, 딥러닝<p /> <p>Language: ko</p>",
language="ko",
issn="1225-6382",
doi="10.7467/KSAE.2022.30.10.839",
url="http://dx.doi.org/10.7467/KSAE.2022.30.10.839"
}