基于大数据的异构矢量图斑地类地块边界特征提取方法  

A method for extracting boundary features of parcels with heterogeneous vector patches based on big data

作　　者：宋佳 SONG Jia(Jiangxi Nuclear Industry Geological Survey Institute,Nanchang,Jiangxi 330038,China)

机构地区：[1]江西省核工业地质调查院,江西南昌330038

出　　处：《北京测绘》2025年第3期358-363,共6页Beijing Surveying and Mapping

基　　金：江西省重点研发计划(20201BBG71001)。

摘　　要：与常规图斑相比,异构矢量图斑的分割难度更大,难以获得其边界特征,因此提取的地块边界特征容易出现不清晰、连续性较差的问题。为此,本文提出基于大数据的异构矢量图斑地类地块边界特征提取方法。在多尺度视网膜皮层(Retinex)算法中引入灰度校正,利用改进后的多尺度Retinex算法,均衡化处理异构矢量斑块。采用大数据技术中的显著图(Itti)模型,提取图斑的显著区域,融合处理异构矢量图斑的显著区域,在显著区域内,将图斑显著区域划分为背景像素集与目标像素集,提取异构矢量图斑的特征映射图。利用Canny算法,提取地类地块边界特征,对应真实的边缘点。实验结果表明,所提方法的均衡化处理结果评价函数值得到提升,显著区域提取结果不存在误提取和漏提取的问题,且地类地块区域边界特征对应的像素点清晰,连续性强。Compared with that of conventional patches,the segmentation difficulty of heterogeneous vector patches is greater,and it is difficult to obtain their boundary features.Therefore,the extracted parcel boundary features are unclear,with poor continuity.This paper proposed a method for extracting boundary features of parcels with heterogeneous vector patches based on big data.The paper introduced grayscale correction into the multi-scale Retinex algorithm and utilized the improved multi-scale Retinex algorithm to balance the processing of heterogeneous vector patches.The Itti model in big data technology was employed,and the salient regions in patches were extracted.In addition,the salient regions in heterogeneous vector patches were fused and processed.Within the salient regions,the salient regions in patches were divided into a background pixel set and a target pixel set,and the feature map of heterogeneous vector patches was extracted.The Canny algorithm was used,and boundary features of parcels were extracted,corresponding to real edge points.The experimental results show that the evaluation function of the balanced processing results of the proposed method is worthy of improvement,and there is no problem of erroneous or missed extraction in salient regions.Moreover,the pixel points corresponding to the boundary features of parcels are clear and have strong continuity.

关 键 词：大数据技术 异构矢量图斑 多尺度视网膜皮层(Retinex)算法 显著区域 CANNY算法 

分 类 号：P232[天文地球—摄影测量与遥感]