Steady corner detection for calibration in underwater environment

Abstract

Meanshift algorithm has been widely used in the fields of image filtering, image segmentation and object tracking. This article tries using Meanshift algorithm in corner detection. For camera calibration in the underwater environment, the article proposed a corner detection method based on Harris corner detection and Meanshift algorithm. First, calculate the Harris corner and the dot product between the shift vector and gradient vector of the underwater chessboard image, and combine the two characteristics to generate the probability density of right-angle corner measure, then take the Harris corner as the initial value, use the Meanshift algorithm to calculate high-precision corner locations and identify the false corners to remove them. The results of experiments show that the algorithm can achieve sub-pixel precision, and is suitable for calibration of underwater camera.