물 자연 그리고 사람 - 물로 더 행복한 세상을 만들어가겠습니다.
HOME성과논문실적

논문실적

Automated Procurement of Training Data for Machine Learning Algorithm on Ship Detection Using AIS Information 게시글의 제목, 학술지명, 저자, 발행일, 작성내용을 보여줌
Automated Procurement of Training Data for Machine Learning Algorithm on Ship Detection Using AIS Information
학술지명 Remote Sensing 저자 송주영,김덕진,강기묵
발표일 2020-05-02

Development of convolutional neural network (CNN) optimized for object detection, led to significant developments in ship detection. Although training data critically affect the performance of the CNN-based training model, previous studies focused mostly on enhancing the architecture of the training model. This study developed a sophisticated and automatic methodology to generate verified and robust training data by employing synthetic aperture radar (SAR) images and automatic identification system (AIS) data. The extraction of training data initiated from interpolating the discretely received AIS positions to the exact position of the ship at the time of image acquisition. The interpolation was conducted by applying a Kalman filter, followed by compensating the Doppler frequency shift. The bounding box for the ship was constructed tightly considering the installation of the AIS equipment and the exact size of the ship. From 18 Sentinel-1 SAR images using a completely automated procedure, 7489 training data were obtained, compared with a di erent set of training data from visual interpretation. The ship detection model trained using the automatic training data obtained 0.7713 of overall detection performance from 3 Sentinel-1 SAR images, which exceeded that of manual training data, evading the artificial structures of harbors and azimuth ambiguity ghost signals from detection.

목록