Optimization of Binocular Vision Ranging Based on Sparse Stereo Matching and Feature Point Extraction

In recent years, the advancement in ranging accuracy and distance requirements in the ship industry has underscored the paramount importance of controlling ranging error. Beyond a reasonable range, any error may result in irreparable losses, including those pertaining to the safety of the ship and its personnel. Therefore, this study proposes a binocular vision ranging method based on sparse stereo matching and feature point extraction. The method accurately extracts the feature points in the image by optimizing the feature extraction algorithm and fast robust feature matching filtering algorithm. The experimental results showed that in data 1, the non-great suppression algorithm suppressed 5111 feature points when the suppression ratio was 20%. In data 2, the improved algorithm suppressed 1920 feature points when the suppression point feature ratio was 70%, while it was 3950 when the ratio was 20%. The total number of extracted feature points was the most in data 3. Moreover, the improved algorithm suppresses 4060 feature points in 70% ratio, compared to 89050 in 20% ratio. The results reveal that the ranging algorithm has higher accuracy in feature point extraction, which provides an important theoretical basis for the intelligent development of the ship industry.