CAJ | 학술논문

传统的QR码识别算法只适用于打印在平面上的条码，提出了一种有效识别打印在饮料瓶等圆柱面上的QR码。通过对图像轮廓进行角点检测确定回字定位图形，在此基础上筛选条码关键轮廓并对其进行霍夫变换提取圆柱面上的透视椭圆信息，同时结合透视椭圆的参数和三维透视变换，有效构建了圆柱面条码像素从二维图像平面直接映射到三维图像空间的变换矩阵，重构打印在平面或圆柱面上的QR码目标。实验结果表明，该算法对平面或圆柱面QR条码的识别有较高的准确率。
전통적QR마식별산법지괄용우타인재평면상적조마，제출료일충유효식별타인재음료병등원주면상적QR마。통과대도상륜곽진행각점검측학정회자정위도형，재차기출상사선조마관건륜곽병대기진행곽부변환제취원주면상적투시타원신식，동시결합투시타원적삼수화삼유투시변환，유효구건료원주면조마상소종이유도상평면직접영사도삼유도상공간적변환구진，중구타인재평면혹원주면상적QR마목표。실험결과표명，해산법대평면혹원주면QR조마적식별유교고적준학솔。
Traditional QR code recognition algorithm is usually applied to the barcode printed on flat surface only. A low-cost approach to recognize the curved QR codes printed on bottles or cans is proposed in this paper. The width proportion and corners of image contours are extracted to confirm the positioning patterns and an efficient Hough transformation ellipse fitting method is employed to extract the elliptic information. In combination with the parameters of perspective ellipse and 3D perspec-tive transformation,the transformation matrix of barcode pixels on a cylindrical surface is constructed by direct mapping from the 2D image plane to 3D image space. The experiment result proves the algorithm has the high-accuracy recognition ability of barcodes no matter on the flat or the cylindrical surface.