摘要
自组织特征映射神经网络SOFM可以实现无监督的特征聚类.利用SOFM实现逆向工程中点云数据分区,通过改进SOFM网络初始权值方法以及引进能量函数控制迭代次数,提高了SOFM的分区效率.利用SOFM方法实现点云数据分区具有较强的容错性能,对测量数据点无任何要求.实例运行结果验证了此方法的可行性.
Self_organizing feature map (SOFM) neural network can implement feature vector clustering without teachers, thus SOFM can be used for segmentation of point cloud on reverse engineering. The efficiency of segmentation is improved by modifying initial weight vectors and adding energy function to control iteration numbers. The segmentation of point cloud using SOFM is robust to noise and has no limitation for data point type. The method is validated by the real scanned point_cloud.
出处
《华北水利水电学院学报》
2004年第2期59-62,共4页
North China Institute of Water Conservancy and Hydroelectric Power
关键词
自组织特征映射
神经网络
数据分区
逆向工程
self-organizing feature map
neural networks
point cloud segmentation
reverse engineering
