基于改进GA优化BP网络的双目视觉定位研究  

Binocular Vision Positioning Research Based on Improved GA Optimized BP Network

作　　者：褚新建 何丹 张伏 CHU Xin-jian;HE Dan;ZHANG Fu(School of Mechanical and Electrical Engineering,Zhengzhou Institute of Industrial Application Technology,Zhengzhou 451100,China;School of Information Engineering,Zhengzhou Institute of Industrial Application Technology,Zhengzhou 451100,China;College of Agricultural Equipment Engineering,Henan University of Science and Technology,Luoyang 471003,China)

机构地区：[1]郑州工业应用技术学院机电工程学院,郑州451100 [2]郑州工业应用技术学院信息工程学院,郑州451100 [3]河南科技大学农业装备工程学院,洛阳471003

出　　处：《组合机床与自动化加工技术》2023年第2期28-30,36,共4页Modular Machine Tool & Automatic Manufacturing Technique

基　　金：河南省高等学校青年骨干教师培养计划(2021GGJS190);教育部高教司产学合作协同育人项目(201801047029)。

摘　　要：为了提高传统遗传算法(genetic algorithm, GA)IGA优化BP网络迭代时间过长以及精度偏低的缺陷,设计了一种通过改进遗传算法(improved genetic algorithm, GA)IGA优化BP网络,并进行完成双目视觉的定位计算。改进遗传算法来提升BP网络收敛能力并获得更强的全局寻优效果,显著改善BP网络处理效率与精度,最终促使相机获得更高定位精度以及运算速率。给出了IGA优化BP网络的双目视觉定位算法流程,并开展了双目视觉定位实验。研究结果表明,未优化坐标预测值误差均值为0.66 mm,优化坐标误差均值为0.08 mm。改进BP网络进行双目视觉定位精度达到0.12 mm,相对最初预测定位误差降低近0.01 mm。以BP网络来定位双目视觉精度均值是0.12 mm,以OpenCV定位的实际精度是0.10 mm。推断以神经网络双目视觉进行定位时满足双目视觉定位精度条件。In order to improve the defects of the traditional genetic algorithm(GA) IGA optimized BP network with excessively long iteration time and low accuracy, a IGA optimized BP network with improved GA was designed.And completes the binocular vision localization calculation.Genetic algorithm was improved to improve the convergence ability of BP network and obtain stronger global optimization effect, significantly improve the processing efficiency and accuracy of BP network, and finally promote the camera to obtain higher positioning accuracy and operation rate.The binocular vision localization algorithm flow of IGA optimized BP network is given, and the binocular vision localization experiment is carried out.The results show that the mean error of unoptimized coordinate predicted value is 0.66 mm, and the mean error of optimized coordinate is 0.08 mm.The improved BP network achieves the positioning accuracy of 0.12 mm, which reduces the positioning error by nearly 0.01 mm compared with the initial prediction.The average accuracy of binocular vision positioning by BP network is 0.12 mm, and the actual accuracy of binocular vision positioning by OpenCV is 0.10 mm.It is concluded that the accuracy condition of binocular vision is satisfied when neural network binocular vision is used for localization.

关 键 词：双目视觉定位 BP网络 改进遗传算法 

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]