中文摘要：

颗粒物质在自然界中普遍存在,如砂砾、谷物、堆石体和碎屑流等,许多自然现象和工业过程都与颗粒物质的运动状态有关。弄清颗粒之间的接触特征是理解颗粒变形与运动规律的关键,与之相关联的接触点确定是获取全场接触力（包括力链）信息的难点。近年来,光弹法逐渐成为这类问题研究的有效手段之一,原因在于该方法具有原理简单、操作方便、图像直观等优点。然而,已有的接触点判定方法主要依赖于圆盘圆心距离的计算,对于小变形圆盘接触特征难以有效识别。为此,本文发展了一种基于去除零级条纹区域及图像矩阵转换的算法,可以有效地确定接触点的位置。主要步骤如下：首先运用主成分分析方法对图像进行分析,提取主成分,并重构原图像;然后利用模板匹配方法提取各个光弹圆盘,接着对提取出来的各个圆盘进行图像矩阵变换;最后对变换后的图像局部进行垂直方向灰度梯度平方计算,描绘梯度变化曲线并提取曲线峰值,从而识别出接触点（位置）。

英文摘要：

Granular material,such as gravel,grain,rockfill and debris flow,is common in nature,and a lot of natural phenomena and industrial processes are related to granular material motion state.Gaining a clear idea of contact characteristics among granules is the key to understand the deformation and movement of granules.Identifying contact points is the difficult point for obtaining the information of full field contact force （including the force chain）.In recent years,photoelastic method has become one of effective means to study this problems due to its simple principle,convenient operation and visual image.However,existing method for contact point identification mainly depends on the calculation of disk center-to-center distance,which is difficult to effectively identify the contact features of disks with smaller deformation.Therefore,a new image matrix transformation algorithm based on zero-order fringe elimination is presented in this paper to effectively identify the contact point location.The main steps are as follows：firstly,principal component analysis is applied to extract principal component of images,and original images are reconstructed;then,template matching method is adopted to extract each disk and image matrix transformation for each extracted disks is carried out;finally,gray gradient square calculation along perpendicular direction is performed for local areas of transformed image,the gradient variation curve is plotted,its peak value is obtained.Thus,contact point locations are identified.