Matching Cross-Domain Data with Cooperative Training of Triplet Networks: A Case Study on Underwater Robotics

• Published in: Journal of intelligent & robotic systems Vol. 104; no. 3; p. 55
• Main Authors: De Giacomo, Giovanni G., dos Santos, Matheus M., Drews-Jr, Paulo L. J., Botelho, Silvia S. C.
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 01.03.2022; Springer; Springer Nature B.V
• Subjects: Acoustics; Algorithms; Artificial Intelligence; Artificial neural networks; Case studies; Control; Domains; Electrical Engineering; Engineering; Localization; Machine vision; Matching; Mechanical Engineering; Mechatronics; Ranking; Rankings; Regular Paper; Robotics; Robotics industry; Robots; Topical collection on LARS 2020; Training; Underwater acoustics; Underwater robots; Deep learning; 68T45; 68T40; Image fusion; Cross-domain matching; Triplet networks
• ISSN: 0921-0296; 1573-0409
• DOI: 10.1007/s10846-022-01591-7

Recently, Deep Convolutional Neural Networks have been successfully applied to various robotics problems, such as robot vision and simultaneous localization and mapping. Among these, siamese and triplet networks have obtained great traction in intra-domain matching. However, it is impossible to directly use these networks in cross-domain problems. Thus, this paper proposes a new method to train a set of triplet networks to perform cross-domain matching and ranking focused in underwater robotics. The method is used to train a pair of networks to perform matching of acoustic and segmented aerial images aiming to support an underwater robot localization algorithm. Results show that the method is able to achieve up to 83% accuracy in matching acoustic and segmented aerial images and up to 85% recall in ranking relevant aerial images given an acoustic image.