TY - GEN
T1 - Loop-closure detection by LiDAR scan re-identification
AU - Peltomäki, Jukka
AU - Ni, Xingyang
AU - Puura, Jussi
AU - Kämäräinen, Joni Kristian
AU - Huttunen, Heikki
N1 - Publisher Copyright:
© 2020 IEEE
PY - 2021
Y1 - 2021
N2 - In this work, loop-closure detection from LiDAR scans is defined as an image re-identification problem. Re-identification is performed by computing Euclidean distances of a query scan to a gallery set of previous scans. The distances are computed in a feature embedding space where the scans are mapped by a convolutional neural network (CNN). The network is trained using the triplet loss training strategy. In our experiments we compare different backbone networks, variants of the triplet loss and generic and LiDAR specific data augmentation techniques. With a realistic indoor dataset the best architecture obtains the mean average precision (mAP) above 0.94.
AB - In this work, loop-closure detection from LiDAR scans is defined as an image re-identification problem. Re-identification is performed by computing Euclidean distances of a query scan to a gallery set of previous scans. The distances are computed in a feature embedding space where the scans are mapped by a convolutional neural network (CNN). The network is trained using the triplet loss training strategy. In our experiments we compare different backbone networks, variants of the triplet loss and generic and LiDAR specific data augmentation techniques. With a realistic indoor dataset the best architecture obtains the mean average precision (mAP) above 0.94.
U2 - 10.1109/ICPR48806.2021.9412843
DO - 10.1109/ICPR48806.2021.9412843
M3 - Conference contribution
AN - SCOPUS:85110535433
T3 - Proceedings - International Conference on Pattern Recognition
SP - 9107
EP - 9114
BT - Proceedings of ICPR 2020 - 25th International Conference on Pattern Recognition
PB - IEEE
T2 - 25th International Conference on Pattern Recognition, ICPR 2020
Y2 - 10 January 2021 through 15 January 2021
ER -