目的 针对半径不等的类圆柱标签的全表面展开问题，提出类圆柱标签等效阶梯柱面展开方法。方法 提取4个视角下图像中的类圆柱标签最小外接矩形，根据提取结果建立位姿估计模型。基于微积分的化曲为直的思想，结合标签的位姿和各视角的轮廓，建立等效阶梯的类圆柱标签3D点云。利用双线性插值法对点云进行渲染，并通过NCC算法进行图像拼接，最终实现半径不等的类圆柱标签全表面展开。结果 类圆柱棋盘格标签仿真模型与某品牌口香糖标签全表面展开实验表明，文中提出的算法较传统算法具有较高的精度，针对1号模型，在x轴、y轴方向上的方差分别为1.37像素和0.58像素。结论 类圆柱标签等效阶梯柱面展开方法可以有效地实现类圆柱标签全表面展开，为后续的标签检测提供基础。

The work aims to propose a method for equivalent ladder cylinder unfolding of quasi-cylinder label image to solve the problem on whole surface unfolding of quasi-cylinder labels with unequal radii. Firstly, the minimum bounding rectangles of quasi-cylinder label of images from four perspectives were extracted. A pose estimation model was established according to the result extracted. Based on the quasi-cyl' idea of turning curve into straight, the equivalent ladder quasi -cylinder label 3D point cloud was established in combination with the label pose and outline from different perspectives. Finally, the method of bilinear interpolation was used to render the point cloud to get quasi-cylinder label unfolding results of unequal radii through image stitching with NCC algorithm. The quasi-cylindrical checkerboard label simulation models and the whole surface unfolding experiment of the chewing gum label showed that the proposed algorithm had higher accuracy than the traditional algorithms. For model 1, the variance in the x-axis and y-axis directions was 1.37 pixels and 0.58 pixels respectively. In conclusion, the equivalent ladder cylinder unfolding of quasi-cylinder label can effectively achieve whole surface unfolding of quasi-cylinder label, providing a good experimental basis for the subsequent label detection.