基于ELM的多结构变形QR码分类校正研究-计算机应用研究.PDF

优先出版 计 算 机 应 用 研 究 第33 卷 基于ELM 的多结构变形QR 码分类校正研究* 许 刚，沈宇超，谈元鹏 (华北电力大学 电气与电子工程学院，北京 102206) 摘 要：针对不同曲面上 QR 码变形多样化、识别率低的问题，考虑到极限学习机(ELM)对大量数据的快速分类能力， 提出了一种基于 ELM 的多结构变形QR 码分类校正算法。在欧式距离量化变形特征后，运用ELM 算法把变形结构分 为平面变形、半曲面变形和全曲面变形三类，并利用不同分类系数改进QR 码坐标透视变换算法，得到校正坐标值。实 验结果表明，此方法不仅提高了QR 码在曲面上的校正准确率，而且通过分类提高了曲面变形和平面变形QR 码的校正 速度。 关键词：QR 码多结构变形；极限学习机；分类校正 中图分类号：TP391.4 Multi-structural QR code recognition based on ELM Xu Gang, Shen Yuchao, Tan Yuanpeng (School of Electrical Electronic Engineering, North China Electric Power University, Beijing 102206, China) Abstract: Since the traditional QR code recognition methods fail to fulfill the recognition task on multi-structural deformation deformed QR code cases. this paper presents a novel QR code recognition algorithm based on extreme learning machine. In this method. Euclidean distance of feature points are calculated as deformation characteristics. And employed Extreme learning machine to classify the deformation type. It determined the transform coefficients of each type. which were used for coordinates mapping. Then obtained real coordinates by perspective mapping with different transform coefficients . thus the code were reconstructed. Experimental results demonstrate that our proposed method holds a satisfactory performance on both of recognition rate and time consuming. Key Words: multi-structural QR code; extreme learning machine; classified correction 但当失真角度变大时准确性会降低。变形 QR 码的四个顶点扭 0 引言 曲程度最明显，且形变类型和角度多样化，如果通过二维透视