摘要
首先介绍了一种利用SOM神经网络对顶点进行聚类的线要素简化算法,该算法以线要素各顶点x,y坐标为输入样本集,经过SOM神经网络的训练,形成对原有顶点的聚类,每个聚类保留一个顶点作为简化后的结果。然后分析该算法存在的一些问题,先假设加入角度和距离两维能改善原算法的效果。最后自主实现算法,并采用相关实验数据加以验证,证明增加SOM维数确实行之有效。
Firstly,a linear features simplification algorithm based on vertexs clustering var SOM neural network was introduced.After the SOM neural network was trained by the x,y coordinates of the linear feature' s vertexs,the vertexs of the linear feature was clustered by the SOM neurons,one vertex was selected in every cluster to build up the simplified linear feature.Analysising some problems existing in the algorithm,it was assumed that the additional two dimensional,the angle and the distance,can improve the effect of the original algorithm.The algorithm was implemented independently and validated by related experimental data,which shows that our assumption is indeed correct.
出处
《测绘科学技术学报》
CSCD
北大核心
2013年第5期525-529,534,共6页
Journal of Geomatics Science and Technology
基金
国家自然科学基金项目(41071288)
关键词
制图综合
线要素简化
顶点聚类
SOM神经网络
弯曲
cartographic generalization
linear features simplification
vertex clustering
SOM neural network
bend