Multibeam Data Processing for Underwater Mapping
| Autor: | Pedro V. Teixeira, John J. Leonard, Franz S. Hover, Michael Kaess |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2018 |
| Předmět: |
0106 biological sciences
Data processing 010505 oceanography Computer science business.industry 010604 marine biology & hydrobiology 01 natural sciences Sonar Thresholding Robustness (computer science) Computer vision Artificial intelligence Underwater business 0105 earth and related environmental sciences Subsea |
| Zdroj: | other univ website IROS |
| Popis: | © 2018 IEEE. From archaeology to the inspection of subsea structures, underwater mapping has become critical to many applications. Because of the balanced trade-off between range and resolution, multibeam sonars are often used as the primary sensor in underwater mapping platforms. These sonars output an image representing the intensity of the received acoustic echos over space, which must be classified into free and occupied regions before range measurements are determined and spatially registered. Most classifiers found in the underwater mapping literature use local thresholding techniques, which are highly sensitive to noise, outliers, and sonar artifacts typically found in these images. In this paper we present an overview of some of the techniques developed in the scope of our work on sonar-based underwater mapping, with the aim of improving map accuracy through better segmentation performance. We also provide experimental results using data collected with a DIDSON imaging sonar that show that these techniques improve both segmentation accuracy and robustness to outliers. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|