Active Scanning Method for AUVs Using Entropy Maps Utilized with Underwater Sonar Measurements
1
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
This paper presents an active scanning method for an autonomous underwater vehicle (AUV) equipped with an imagesonar to generate an underwater 3-dimensional (3D) map. As sonar information measured by AUVs is a useful feature to generateaccurate 3D maps underwater, where high turbidity and signal attenuation are prevalent, we propose a two-stage seabed scanningmethod: 1) scan over a designated area to generate a 3D occupancy map and represent uncertain regions in the occupancy map interms of entropy, and 2) find the most uncertain region with the highest entropy in the generated map and search for consecutivescan paths that can reduce the uncertainty of the map the most. The proposed scanning method was tested through computer simulations,in which the experimental settings (AUV, sensors, objects, and measurement conditions) were configured to be similar tothose in the real world. The proposed method was compared with conventional lawnmower-based and uniform scanning methodsby analyzing the precision, recall, and F-scores of the object shapes in the generated underwater maps. The results demonstrate thatthe proposed method can provide efficient scan paths and generate accurate underwater maps.
Abstract
This paper presents an active scanning method for an autonomous underwater vehicle (AUV) equipped with an imagesonar to generate an underwater 3-dimensional (3D) map. As sonar information measured by AUVs is a useful feature to generateaccurate 3D maps underwater, where high turbidity and signal attenuation are prevalent, we propose a two-stage seabed scanningmethod: 1) scan over a designated area to generate a 3D occupancy map and represent uncertain regions in the occupancy map interms of entropy, and 2) find the most uncertain region with the highest entropy in the generated map and search for consecutivescan paths that can reduce the uncertainty of the map the most. The proposed scanning method was tested through computer simulations,in which the experimental settings (AUV, sensors, objects, and measurement conditions) were configured to be similar tothose in the real world. The proposed method was compared with conventional lawnmower-based and uniform scanning methodsby analyzing the precision, recall, and F-scores of the object shapes in the generated underwater maps. The results demonstrate thatthe proposed method can provide efficient scan paths and generate accurate underwater maps.