Dual-frequency acoustic camera: a candidate for an obstacle avoidance, gap-filler, and identification sensor for untethered underwater vehicles

摘要

The Dual-Frequency Identification Sonar (DIDSON) is a forward-looking sonar that can mount on an untethered underwater vehicle (UUV). It performs three important tasks. In the low-frequency mode, it ensonifies the gap between the coverage of two side-scan sonars during surveys and can serve as an obstacle avoidance sonar. In the high-frequency mode, its very high resolution allows the identification of objects in turbid water where optical systems fail. The sonar is small, light, and requires only 30 watts to operate. DIDSON currently is used on three UUVs (two swimmers and one crawler) as part of the Office of Naval Research Undersea, Autonomous Operation Capabilities Program. DIDSON has a 29/spl deg/ field of view and operates at either 1.0 MHz or 1.8 MHz. The Woods Hole REMUS vehicle, in its dual side-scan sonar configuration, has a 6-m to 8-m gap in its coverage. This gap is filled by DIDSON when looking down-range at distances greater than 16 m. The Bluefin Robotics UUV operated by the Coastal Systems Station swims in deeper water, flies higher off the bottom and has a side-scan gap up to 20 m wide. A modified DIDSON that operates at 750 kHz (DIDSON-LR) is proposed for this application. It should image at ranges in excess of 40 m. When operating as a gap-filler, DIDSON collects data at a constant frame rate and stores that data during the duration of the mission. An analysis application is being written to sift through the gigabytes of stored data, locate objects on the seafloor and score them with respect to their mine-like characteristics. Operation efficiency will dramatically increase when UUVs can identify mines autonomously and act upon these identifications. Algorithms are being developed to perform this autonomous identification. The process starts with image processing to extract salient object features. The current approach compares these features to a knowledge base of object features, allowing for object rotation and interaction with the environment. Intelligent algorithms will be developed to associate the object under consideration to objects in the knowledge base in a statistically significant way.