基于全连接神经网络的基础测绘实体分类方法

Basic surveying and mapping entity classification based on fully connected neural network

由于基础测绘实体包含建筑物、道路、植被、水系等多种类型,因此在分类时需采集的数据较多。为了达到理想的分类效果,本文提出基于全连接神经网络的基础测绘实体分类方法。首先,通过基础测绘实体采集融合多源数据,实现不同层次地理实体的数据集成;其次,基于全连接神经网络检测实体影像边缘,将边缘构成特征线,划分不同类型实体特征;再次,分割实体结构提取特征,使属性相同的点从属于相同的实体结构;最后,建立基础测绘实体分类模型,定义各类别语义准则,输出分类结果。试验结果表明,本文所提方法具有较好的分割效果,能提取可区分度较低的特征,在基础测绘实体分类中优势明显,具有一定的可行性。

Basic surveying and mapping entities include various types,such as buildings,roads,vegetation,water systems and many others,so there are a large amount of data needs to be collected during classification.In order to get reasonable classification,a method based on fully connected neural networks is proposed in this paper.Firstly,multi sourced data is integrated through the collection of basic surveying and mapping entities to realize geographical entities integration at different levels.Secondly,the edges of the solid image are detected,and the feature lines are formed based on the fully connected neural network to divide the different features of the entities.Thirdly,the entity structure is segmented to extract features so that the points with the same attributes belong to the same entity structure.Finally,the basic surveying and mapping entity classification model is established,the semantic criteria of each category is defined,and the classification results are output.The experimental results show that the proposed method has good segmentation effect and can extract features with low distinguishability,the F1 index is above 0.85,it has great advantages,and can be used in the classification of basic surveying and mapping entities.