The Research Of Underwater Stereo Matching Algorithms Based On The Binocular Vision

Ocean has abundant resources, and it’s conductive to solve the problem of resources scarcity in land and to keep a sustainable development for human that developing the marine science and utilizing the marine resources reasonably. Vision technology becomes an important underwater detection means in a short distance due to its large advantages, such as non-contact, high precision and large amount of information, and is usually equipped with underwater robots. This paper conducts systematic researches on underwater binocular stereo vision, and applies the algorithm of region matching into underwater environment. Main work and study are as follows(1) The paper starts with the cameral liner projecting model in the air and analyzes the model underwater. Meanwhile, introduces the parallel model in binocular stereo vision, and establishes underwater positioning model considering of refract. The constraints of disparity, basic steps of stereo matching and the classification of stereo matching algorithms are introduced as well.(2) In terms of underwater binocular image matching cannot satisfy the epipolar constraint of air, and the large amount of calculation of underwater image processed by the NCC algorithm, this paper presents an underwater region matching algorithm based on optimum searching area. First, binocular camera should be calibrated in order to obtain some relevant parameters, as well as reference image and image to be matched; then the maximum deviate value from the line in the air can be calculated through the curve constraint and the optimum searching area is therefore decided. Next, the NCC region matching algorithm can help to match two images, at the same time, instead of searching on the original epipolar line; we propose an optimum searching area so that the searching is performed in this area with several lines to achieve the purpose of a higher accuracy. Meanwhile, the time spent on the matching is reduced by the application of box filter technology. The results of the test indicate this algorithm achieves the same matching accuracy compared with the SIFT feature matching algorithm and this can be used to perform dense disparity. Also the speed of matching is largely accelerated compared with the original NCC algorithm. Therefore the region matching algorithm is successfully applied to underwater image matching.(3) In the light of the matching process is affected by the imaging offset caused by waterproof glass of cameral, and the speed of region algorithms is slow, this paper presents a multiple-media stereo matching algorithm based on curve constraint. First, taking thickness of waterproof glass into consideration, a new curve constraint is proposed. Meanwhile, the searching is performed along the curve constraint so that the underwater epipolar constraint is successfully supplied. The matching accuracy can increase by about 3.2% and 5.3% compared with SIFT and NCC algorithms, and the speed of matching is largely accelerated.