Method for Detecting and Ranging an Underwater Guided Light Source Based on Binocular Vision

To meet the requirements of high accuracy, real-time performance, and robustness in autonomous underwater vehicle(AUV) underwater docking, this study proposes a binocular vision-based underwater guidance light source detection and ranging method. The main processes of this method includes underwater camera calibration, denoising of original images, detection of light sources, and location calculation. In the original image denoising stage, a Laplace operator is introduced to improve the mean denoising algorithm and enhance the image to highlight the light source. Then, an adaptive threshold binary image based on the dichotomy is used to effectively separate the light source from the background. Light sources on the left and right imaging planes are then detected and matched. Finally, according to the principle of binocular location, the relative position is solved by using the detected light source information. In this method, traditional mean denoising is improved for an underwater environment, and information of the guided light source is highlighted. The best image binary threshold is obtained by using the connected domain as the cyclic judgment condition, and the light source center is determined by weighted centroid detection method. The method is verified in specific experiments that meet the requirements of underwater docking accuracy. In addition, real-time performance is optimized.