基于边缘增强与视差填充的SGBM视觉测距方法

打开文本图片集

中图分类号：TP391.4;TN911.73 文献标志码：A 文章编号：1671-8755（2025）02-0086-08

Abstract：Because the traditional SGBM（Semi-global Block Matching） stereo matching algorithm does not fully consider the edge contour information，there are holes and noise in the generated parallx map. A parallax filling and noise suppression method based on edge enhancement was proposed to improve the traditional SGBM algorithm.On the basis of using four-direction Sobel operator to enhance the edge information，the least squares interpolation method was used to fillthe holes in the parallx map，and the weighted least squares （WLS） filter was used to further optimize the processing noise，so as to improve thequality and accuracy of the parallax map. Then by performing coordinate mapping of the pixel points of the region of interest extracted by YOLOv5s model，the distance value of the target object was accurately calculated.The experimental results show that compared with the traditional method，the average mismatching rate is reduced by 71.47% ，the target recognition accuracy reaches 93.9% within 200cm ， and the average ranging error rate is only 1.44% . Compared with the traditional method，this method has a great improvement in stereo matching effect and ranging accuracy.

Keywords： Visual ranging method；SGBM algorithm； Edge enhancement； Parallax filing；Noise sup-pression

目标物测距是当前计算机视觉系统中最为关键的任务之一，也是无人机定位系统、医疗应用等多种实际场景的基础[1]。（剩余7249字）