中文摘要：

由于不规则扁平粒状物表面为不规则曲面,且体积小,轮廓不规则,通过二维信息对其进行判断存在一定的局限性,而用常规的物理方法和间接测量方法难以满足有效辨析分选所需的环境、标准设定等多项工艺和工程上的要求。为此,本文提出一种自动识别、分选不规则扁平粒状物的方法。该方法通过二维图像进行三维信息重构,然后利用三维信息选择合适的分选特征值,根据特征值进行分选。该方法运用GPU对程序进行加速,可满足不规则扁平粒状物自动化分选的要求。

英文摘要：

Because the rough-granular object is rather small and its surface is not smooth,there are some limits to judging them through the plane features.The conventional physical methods or indirect measurement methods can not meet the environment,parameter setting and many other processes demands of automatic sorting.This paper proposes an automatic method to sort the rough-granular object.Firstly,the three-dimensional information is reconstructed from images and then the information is applied to select the appropriate eigenvalues.Then the rough-granular object is sorted by the selected eigenvalues.This method makes use of the power of GPU to accelerate the process.The test shows that it can meet the automate-sorting requirements.