Research on Binocular Stereo Vision Calibration Technology Based on Sea Surface Images

The limited texture information and low matching accuracy of sea surface images result in large errors and poor stability when using existing camera external calibration methods for sea surface calibration. By improving it, the accuracy and stability of external parameter calibration have been effectively improved. After obtaining the binocular image, use Wallis filtering to preprocess the image to enhance texture features, and then use the improved AKAZE feature point extraction algorithm for uniform feature point extraction. Use KNN combined with NNDR matching method for coarse matching, and then use RANSAC iterative algorithm for matching purification. Use beam adjustment to optimize camera external parameters for the purified matching points. 1600 feature points were uniformly extracted from the weakly textured sea surface image pairs using measured data. After rough matching, 560 matching points were obtained. RANSAC iteration eliminated coarse points, and ultimately retained 378 optimal matching points. The beam adjustment optimization was used to obtain accurate camera external parameters, verifying the feasibility of the proposed method. This study solves the problems of low accuracy and poor stability in extrinsic calibration of sea scenes, improves the reliability of stereo photography wave measurement technology, and is conducive to its promotion and application.