Study On Stabilizing Method Of UUV Visual Images

While UUV (Unmanned Underwater Vehicle, UUV) is working near the seabed, its needs good situational awareness so that one of the key must have the ability, in which sound and visual light perception vision system is UUV external environment two important means and the optical vision system capable of providing relatively complete target information and guidance UUV accurately identify and locate targets. In the visual aspects of the measurement image stabilization aspects, such as the visual image stable vision measurement can not be ignored important research, it helps to suppress jitter operation target image by UUV for different reasons caused, improve the output image quality.First, I introduce the details of the UUV hardware optical vision system configuration(including underwater cameras, frame grabbers, underwater lighting) and fit them on the program in the UUV. Then I introduce the conversion relationship underwater camera imaging model and several commonly used coordinate systems between them, while the cause blurred image sequence is analyzed and classified, then model and its transformation model for single-frame images. Finally, the image stabilization process in motion estimation and motion vector principle several methods for solving basic aspects were analyzed, in order to enhance the stability of the processed image, I introduce several common movement filtering methods.Secondly, the image stabilization method based on the visual image UUV global information were studied. First, since the imaging environment between underwater environment and the air is different, so in order for the image of underwater noise, serious gray effect and ambiguity, the smoothing filter design and use methods such as histogram equalization, enhanced underwater images contrast, image detail is more obvious, easy to carry out follow-up image feature match. The random vibration of the video image is divided into two categories, for the slow vibration image proposed block matching method, and obtaining the minimum MAD worth to match point match criteria according to SAD, but also with different search paths, they are divided into FS method, TSS method and DS method algorithm,and they can be found by comparing the advantages and disadvantages between them. Then, for the rapid vibration of the image, the algorithm is given for the projection of the image related to the ranks of the value of the gradation information, and put forward relevant curve extreme value search method.Again, light visual image information for local features, image stabilization method is proposed based on image feature information. First, an analysis of the characteristic amount may include underwater targets, including linear, equal amounts of the edge points and the feature, at the same time gives a description of the form thereof. Then put forward for image stabilization feature points algorithm, in order to strike a feature point using the Harris operator on the corner is detected. Finally feature descriptor generating feature vectors, while using the RANSAC algorithm to eliminate between feature points mismatching, the last point for solving inter-image motion vector based on the matched features.Finally, in order to simulate the real underwater environment.We build a test platform in the tank to complete the target image acquisition, image stabilization and completed the test image sequence in the preceding chapters algorithm, the results obtained in the inter image drawn motion vector curve and gives a description of the image stabilization effect and PSNR value as a measure of MAD,eventually drawing a more typical image stabilization algorithm MAD PSNR curve and verify the effectiveness of the algorithm.