基于改进SGM算法的双目立体匹配  

Binocular Stereo Matching Based on Improved SGM Algorithm

作　　者：宋坤 易怀安 舒爱华 宋欣茹 黄杰锋 SONG Kun;YI Huaian;SHU Aihua;SONG Xinru;HUANG Jiefeng(School of Mechanical and Control Engineering,Guilin University of Technology,Guilin 541006,China;School of Computer and Information Engineering,Fuyang Normal University,Fuyang 236000,China)

机构地区：[1]桂林理工大学机械与控制工程学院,桂林541006 [2]阜阳师范大学计算机与信息工程,阜阳236000

出　　处：《组合机床与自动化加工技术》2023年第12期142-146,共5页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：国家自然科学基金资助项目(52065016)。

摘　　要：SGM算法在立体匹配中对遮挡区域及视差不连续区域匹配能力较弱,容易产生较多空洞。针对此问题,提出一种SGM算法结合泛洪填充和中值平滑的视差优化算法,该算法是根据空洞边缘的视差值对缺失的视差值进行填充,即可实现对视差图中缺失视差的修复,并把所提算法在Middlebury立体匹配数据集上利用均方根误差(RMSE)、峰值信噪比(PSNR)、结构相似性(MSSIM)三个评价指标进行定量分析。实验结果表明对SGM算法的改进和优化能够有效地填充视差图中的空洞,减少图像的噪声,且视差图的匹配精度比SGM平均提高9.00%。The SGM algorithm is weak in stereo matched to the occluded area and parallax discontinuity area.Tend to produce more voids,In response to this question.An SGM algorithm combines flood fills and median smooths for parallax optimization is proposed.The algorithm is to fill the missing parallax values based on the parallax values at the edge of the cavity.without the need to raise a larger computational volume.the repair of missing parallax in the parallax map can be achieved.The proposed algorithm is also quantitatively analyzed on the Middlebury stereo matching dataset using three evaluation metrics:root mean square error(RMSE),peak signal-to-noise ratio(PSNR),and structural similarity(MSSIM).The experimental results show that,The improvement and optimization of SGM algorithm can effectively fill the voids in the parallax map.reduce the noise of the image,And parallax map matching accuracy is on average 9%better than SGM.

关 键 词：SGM 立体匹配 中值平滑 泛洪填充 

分 类 号：TH165[机械工程—机械制造及自动化] TG659[金属学及工艺—金属切削加工及机床]