사이드스캔 소나 영상에서 물체 인식을 위한 특징점 기술자 결정 방법 연구
Keypoint Descriptor Decision Methods for Object Recognition in Sidescan Sonar Images
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
In side-scan sonar images, object shapes are often deformed by operating conditions such as pulse repetition interval (PRI), ship velocity, and redirection, which are generally expressed as affine transforms. While the scale-invariant feature transform (SIFT) and its variants have been developed to improve scale, rotation, and affine variance, they remain weak in handling shearing problems in sonar fields. To address this issue, we propose an extraction method for robust keypoint descriptors in the shearing form. The proposed method is based on operating conditions, and it reflects the characteristics of vertical and horizontal resolutions. It was generated based on the ship velocity, PRI, and pulse length changes. In addition, sonar motion was used to express the distortion phenomenon. For descriptor extraction, effective keypoints compared the distortion characteristics with the existing SIFT descriptor for shearing and sorting of the descriptors. Experimental results from a sensitivity analysis showed that affine images were particularly sensitive to the sheared images. The proposed method demonstrated better accuracy rate compared to that of conventional methods.
Abstract
In side-scan sonar images, object shapes are often deformed by operating conditions such as pulse repetition interval (PRI), ship velocity, and redirection, which are generally expressed as affine transforms. While the scale-invariant feature transform (SIFT) and its variants have been developed to improve scale, rotation, and affine variance, they remain weak in handling shearing problems in sonar fields. To address this issue, we propose an extraction method for robust keypoint descriptors in the shearing form. The proposed method is based on operating conditions, and it reflects the characteristics of vertical and horizontal resolutions. It was generated based on the ship velocity, PRI, and pulse length changes. In addition, sonar motion was used to express the distortion phenomenon. For descriptor extraction, effective keypoints compared the distortion characteristics with the existing SIFT descriptor for shearing and sorting of the descriptors. Experimental results from a sensitivity analysis showed that affine images were particularly sensitive to the sheared images. The proposed method demonstrated better accuracy rate compared to that of conventional methods.
참고문헌 (8)
- Langner, F., "Image processing in side scan sonar images for object detection and classification", Proceedings of the 3rd International Conference & Exhibition on Underwater Acoustic Measurements, pp.311~318 (2009)
- Klaucke I., "Sidescan sonar in Submarine Geomorphology", pp.13~24 (2018), Springer International Publishing
- Lowe, D. G., "Distinctive Image Features from Scale-Invariant Keypoints", International Journal of Computer Vision, 60, pp.91~110 (2004)
- L. Seidenari, "Local Pyramidal Descriptors for Image Recognition", IEEE Transactions on Pattern Analysis and Machine Intelligence, 30(5), pp.1033~1040 (2014)
- Morel, Jean-Michel, "ASIFT : A new framework for fully affine invariant image comparison", SIAM Journal on Imaging Sciences, 2(2), pp.438~469 (2009)
- The Effects of Towfish Motion on Sidescan Sonar Images, "Anstee, S. D.", Department of Defense DSTO Materials Research Laboratory, ~ (1994)
- Danielsson, P. E., "Euclidean distance mapping", Computer Graphics and Image Processing, 14(3), pp.227~248 (1980)
- Zhang, C. T., "Recognition of protein coding genes in the yeast genome at better than 95%accuracy based on the Z curve", Nucleic acids research, 28(14), pp.2804~2814 (2000)