[ISMT2021] Object detection in high-resolution sonar images
1
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
In this study, object recognition and detection in sound navigation and ranging (sonar) images are investigated using the You Only Look Once approach. Small objects in large images and sparse data pose difficulties to object detection in sonar images. To solve the former problem, an image-segmentation method is proposed herein. Segmentation partitions a large image into smaller sec-tions and then recombines them after object detection is performed. Because object detections in smaller images are performed in parallel, the processing time does not increase significantly. Meanwhile, to solve the latter problem, a large-scale labeled dataset (60,000 images comprising nine classes: tire, diver, shelter, ladder, frame, drum, bedrock, pier, and sandbar) is built by applying various data augmentation methods (reflection, rotation, distortion, etc.). The proposed method shows favorable object recognition and detec-tion performances, with a mean average precision score of 0.745 for 6,000 test data points. In the future, we plan to improve the performance by optimizing the hyperparameters and applying noise-reduction techniques at the preprocessing stage.
Abstract
In this study, object recognition and detection in sound navigation and ranging (sonar) images are investigated using the You Only Look Once approach. Small objects in large images and sparse data pose difficulties to object detection in sonar images. To solve the former problem, an image-segmentation method is proposed herein. Segmentation partitions a large image into smaller sec-tions and then recombines them after object detection is performed. Because object detections in smaller images are performed in parallel, the processing time does not increase significantly. Meanwhile, to solve the latter problem, a large-scale labeled dataset (60,000 images comprising nine classes: tire, diver, shelter, ladder, frame, drum, bedrock, pier, and sandbar) is built by applying various data augmentation methods (reflection, rotation, distortion, etc.). The proposed method shows favorable object recognition and detec-tion performances, with a mean average precision score of 0.745 for 6,000 test data points. In the future, we plan to improve the performance by optimizing the hyperparameters and applying noise-reduction techniques at the preprocessing stage.
참고문헌 (24)
- 박재홍, "하이브리드 판형 열교환기 적용 컴팩트 스팀 유닛 개발을 위한 성능 평가에 관한 실험적 연구", 한국마린엔지니어링학회지, 41(4), pp.296~301 (2017), 한국마린엔지니어링학회[DOI]
- Z. Zou, "Object detection in 20years: a survey", IEEE Computer Vision and Pattern Recognition, 2 (2019)
- S. Reed, "Automated approach to classification of mine-like objects in side scan sonar using highlight and shadow information", IEE Proceedings-Radar, Sonar and Navigation, 151(1), pp.48~56 (2004)
- S. Kim, "The reason why to use acoustic waves on the seabottom survey", Journal of the Korean Society of Marine Engineering, 32(4), pp.481~489 (2008)
- L. Jian, "Active object detection in sonar images", IEEE Access, 8, pp.102540~102553 (2020)
- J. Redmon, "You only look once: unified, real-time object detection", Computer Vision and Pattern Recognition, pp.779~788 (2016)
- G. J. Dobeck, "Automated detection and classification of sea mines in sonar imagery", Proceeding of SPIE, 3079, pp.90~110 (1997)
- M. E. Clarke, "Using the seabed AUV to assess populations of groundfish in untrawlable areas", The Future of Fisheries Science in North America, pp.357~372 (2009)
- M. Valdenegro-Toro, "Object recognition in forward-looking sonar images with convolutional neural networks", Proceedings of OCEANS 2016 MTS, 2016, 1~6
- J. Kim, "The convolution neural network based agent vehicle detection using forward-looking sonar image", Proceedings of OCCEANS 2016 MTS, 2016, 1~5
- H. Baek, "2d and 3d mapping of underwater substructure using scanning SONAR system", Journal of Korea society for Naval Science and Technology, 2(1), pp.21~27 (2019)
- B. Settles, "Synthesis Lectures on Artificial Intelligence Machine Learning, vol. 6", pp.1~114 (2012)
- S. Vijayanarasimhan, "Large-scale live active learning: Training object detectors with crawled data and crows", International journal of computer vision, 108(1-2), pp.97~114 (2014)
- S. Ren, "Faster R-CNN : towards realtime object detection with region proposal networks", Advances in neural information processing system, 28, pp.91~99 (2015)
- L. Henriksen, "Real-time underwater object detection based on an electrically scanned high-resolution sonar", Proceedings of IEEE Symposium on Autonomous Underwater Vehicle Technology, pp.99~104 (1994)
- J. W. Kaeli, "Real-time anomaly detection in side-scan sonar imagery for adaptive AUV missions", Proceedings of IEEE, 2016, 85~89
- E. Galceran, "A real-time underwater object detection algorithm for multi-beam forward looking sonar", Proceedings of International Federation of Automatic Control, 45(5), pp.306~311 (2012)
- D. Zhou, "IoU loss for 2d", 3d object detection, 2019, 85~94
- "Github"[URL]
- S. Cheng, "Abnormal water quality monitoring based on visual sensing of three-dimensional motion behavior of fish", Symmetry, 11(9) (2019)
- C. Shorten, "A survey on image data augmentation for deep learning", Journal of Big Data, 6(1) (2019)
- "Github"[URL]
- 강현철, "소나영상을 이용한 수중 물체의 식별", 전자공학회논문지, 53(3), pp.91~98 (2016), 대한전자공학회
- 김정문, "사이드 스캔 소나 영상에서 수중물체 자동 탐지를 위한컨볼루션 신경망 기법 적용", 한국음향학회지, 37(2), pp.118~128 (2018), 한국음향학회[DOI]