Abstract: After analyzing the characteristic of solid propellants grains crack image and the limitations of traditional edge detection methods，Beamlet analysis，a new multiscale geometry tool for higher-dimension singularities，is applied in crack edge detection of grains image.Firstly，crack image is segmented to extract crack region.Then the segmented image is decomposed by quad-tree to get the optimal image dyadic square.Finally every image square is transformed by Wedgelet to pick out the optimal Beamlets as the edges.Experiments show that the method can not only detect edges in crack image as the traditional methods but also provide more real-time geometric information，like location，length，orientation，which will help to study structural integrity of propellants grains and provide fundamental base for the computer aided analysis.

Key words: Beamlet, Wedgelet, edge detection, propellants grains

摘要： 在分析某固体火箭发动机推进剂药柱裂纹图像自身特点和传统边缘检测方法局限性的基础上，将Beamlet高维奇异几何分析工具应用于裂纹图像的边缘检测中，并给出线段的长度和斜率的Beamlet基链的近似计算公式。首先对图像进行分割，以提取裂纹区域；然后对已分割的图像进行四叉树分解，以最佳划分图像；最后对图像子块进行Wedgelet变换，筛选出最优Beamlet基作为边缘显示出来。结果表明，能够在检测出裂纹边缘的同时给出边缘的位置、长度、方向等更具体的几何信息，这将为药柱结构完整性研究和药柱裂纹图像计算机辅助分析打下基础。

关键词: Beamlet, Wedgelet, 边缘检测, 推进剂药柱