基于Kohonen神经网络的深度图像分割方法

提出一种基于Ｋｏｈｏｎｅｎ神经网络无监督的深度图像分割方法。首先计算了深度图像各点的导数，进而得到各点的法向，以法向和深度值作为每一点的特征矢量，引入自组织神经网络进行初始的聚类；为消除初始聚类产生的过分割现象，采取相邻表面片法向分析的方法进行再分割，得到最终的分割结果。本方法避免了通常的区域分割方法初始种子不易选取的弱点，聚类所用样本少，速度快。实验结果表现了算法良好的性能。

Segmentation of range image based on Kohonen neural network

In this paper, a unsupervised range image segmentation based on Kohonen neural network is presented. At first, the derivative and partial derivative of each point are calculated and the normals in each point are got. With the character vector including normal and range value, self-organization map is introduced to cluster. The normal analysis is used to eliminate over-segmentation and the last result is got. This method avoids selecting original seeds and uses fewer samples, moreover computes rapidly. The experiment result shows that it has a better performance.