基于非极小值抑制的双目视觉测量优化方法  

The Optimization of Binocular Vision Measurement Based on Non-Minimal Suppression

作　　者：欧阳俊祎 许德章[1,2] OUYANG Junyi;XU Dezhang(The College of Artificial Intelligence,Anhui Polytechnic University,Wuhu 241000,China;Wuhu Anpu Robotics Industry Technology Research Institute Co.,Ltd,Wuhu 241000,China)

机构地区：[1]安徽工程大学人工智能学院,安徽芜湖241000 [2]芜湖安普机器人产业技术研究院有限公司,安徽芜湖241000

出　　处：《东莞理工学院学报》2024年第5期91-96,共6页Journal of Dongguan University of Technology

基　　金：国家自然科学基金(52005003);芜湖市科技计划项目(2022jc41);安徽高校协同创新项目(GXXT2023076)。

摘　　要：视觉测量精度受限于系统的固有误差。提出一种基于RANSAC非极小值误差抑制模型,以提高双目视觉测量精度并增强系统鲁棒性。其在传统双目测量模型基础上,引入最小二乘误差评价函数,建立空间参考点以量化测量误差,并以所提误差评价函数作为目标函数,利用RANSAC迭代优化初始测量模型,进而得出总误差最优模型。实验结果表明,较之传统模型,其最大误差由0.359 mm降低至0.200 mm,平均误差由0.186 mm降低至0.115 mm,误差抑制率达38.2%。所提方法能有效改善双目测量系统精度和可靠性,在视觉精确测量领域具有广泛应用前景。The accuracy of visual measurements is limited by inherent errors within the system.A novel approach based on the RANSAC non-minimal error suppression model was proposed to enhance the precision and robustness of binocular visual measurements.Building upon the traditional binocular measurement model,a least squares error evaluation function was introduced to quantify the measurement errors using spatial reference points.The proposed error evaluation function was employed as the objective function for model optimization through iterative RANSAC processes,resulting in the optimal model with minimum total error.Experimental results demonstrated that compared to the traditional model,the proposed method achieved a reduction in maximum error from 0.359 mm to 0.200 mm,and an average error decreased from 0.186 mm to 0.115 mm.This corresponded to an error suppression rate of 38.2%.The proposed method can improve the accuracy and reliability of binocular measurement system,and holds significant potential for widespread applications in the field of precise visual measurements.

关 键 词：双目视觉 RANSAC 误差抑制 模型优化 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]