一种曲面药瓶标签校验方法关键技术

针对药瓶标签具有曲面性以及采集设备视角小的问题,提出了一种基于全景拼接技术的药瓶标签校验方法.采用柱面反投影算法将发生畸变的曲面药瓶标签进行校正,使用SIFT尺度不变特征算法对校正后标签图像进行拼接,提取拼接后的标签名等信息,采用基于连通域的字符分割技术进行字符切分,计算分割字符与模板字符的特征矩阵及相似度矩阵,从而识别出药瓶标签.结果表明,该方法能够校正标签两侧的文字压缩,实现图像拼接,得到包含完整药瓶名称的图像,并能够实现字符的准确切分,具有速度快和识别率高的优点.

Key technology of checking method for curved surface label of medicine bottle

Aiming at the problem that the label of medicine bottle has the curved surface and the viewing angle of collection device is small, a checking method for the label of medicine bottle based on the panoramic stitching technique was proposed. The distorted curved surface label of medicine bottle was corrected with the cylindrical reverse projection algorithm, and the corrected label of medicine bottle was stitched with the SIFT scale invariant feature transform algorithm. And the label name of medicine bottle after stitching was extracteed. The character segmentation technique based on the connected domain was used to segment the characters, and the feature matrix and similarity matrix of both segmented characters and template characters were calculated to check the label of medicine bottle. The results show that the proposed method can correct the text compression on both sides of the label, realize the image stitching, get the image containing the complete bottle name, achieve the accutate character segmentation, and has the advantages of fast speed and high recongnition rate.