Research On Vision Guided Technology In Piggyback Recovery For UUV

As the competition of oceanic resources becomes more and more intense, UnmannedUnderwater Vehicles (UUVs), which is one of the important instruments for oceanicexploration, have become a popular research topic for many research institutions anduniversities. Also UUVs have become an important military equipment need to be developedwith lots of intention. Underwater recovery of UUV is a key technology to guaranteecontinuity and secrecy of undersea tasks, so the piggyback recovery method has receivedmore and more adorations because of its excellent privacies and flexibilities. Docking withthe submarine is a crucial step for piggyback recovery with a strict request for the locatingaccuracy. Comparing with other sensors, vision sensors could not only supply sufficient visualinformation of environment, but also has much higher positioning accuracy in short distance.However, low underwater visibility, plenty kinds of biologics and protean environment makeapplication of undersea vision technology a challengeable research project. In order to solvethose problems, this paper aims at investigating and developing a single camera visionguiding and locating system for UUV undersea piggyback recovery with sufficientpositioning accuracy.A linear guided lights system, which includes one reference light and eight ordinarylights, is adopted to satisfy UUV’s (the camera is fixed in the middle of UUV’s belly) requestfor acquisition of the submarine’s sway, surge, heading and height information. The referencelight is used to identify the submarine’s heading orientation, and the other eight ordinarylights are designed to locate the submarine’s sway, surge and height. To eliminate impacts onpositioning accuracy, a camera virtual rotation approach is adopted to compensate UUV’s rollor pitch during the docking process.Whether or not the lights areas are extracted accurately is a precondition of calculatinglocating information precisely. Segmenting images according to improved planar Otsualgorithm, which is much more accurate and faster than traditional Otsu algorithm, to achieveoriginal light areas is adopted in this paper. However, fake light areas may appear due to thecomplex underwater environment, then Blob analyses are used to describe all original lightareas and the fake ones are elimated according to Byes Decision Rules. Finally, MathematicalMorphology operators are adopted to modify the lights areas obtained in order to makeoutlines slippery so the character points can be obtained easily. As the UUV and the submarine are both moving during the recovery procedure, weshould acquire their motion information as well as position and attitude datas. Besides, greatlychanging backgrounds of continuous images make it difficult to implement frame differencemethod, background subtractions method and so on. The optical flow method, however,doesn’t require any information about the backgrounds when calculating velocities of movingobjects. Hence, the optical flow method is more adaptive to UUV’s piggyback recovery thanother solutions. Nevertheless, it is quite complicated to calculate whole images’ optical flowfield, so another optical flow calculating method based on character points matching isadopted in this paper. This method could decrease time and space complexities to a greatextent and satisfy real-time demands for recovery. After matching character points ofcontinuous images in video sequences and calculating the optical flow field, the relativevelocities and attitude information of UUV and the submarine’s are obtained. Then the worldcoordinate position and velocity of submarine can be acquired through coordinatetransformations.Finally, simulation experiments are implemented in usage of the11-DOF platform, amonocular camera and the lights guidance system. Line motion, rotative motion andcomplicated curvilinear motion recovery procedures are designed according to actual tasks.Results show that guiding lights abstraction and motion detection methods are effective andthe monocular vision guided positioning system is executable.