首页|知识引导的碎片化栅格地形图比例尺智能识别

知识引导的碎片化栅格地形图比例尺智能识别

扫码查看
原文链接
  • 万方数据
  • 维普
比例尺是确定地形图秘密等级的重要依据.本文针对碎片化栅格地形图比例尺判定的难题,通过凝练地图尺度特征先验知识,引导构建专家知识图像金字塔数据集(EKIPD),然后使用深度卷积神经网络算法进行建模,构建以知识为引导,以数据为驱动,以算法为核心的知识、数据与深度卷积神经网络耦合的混合智能模型.统计EKIPD中不同尺寸碎片化地形图的样本分布得到最优识别尺寸(ORS),然后以ORS为步长对待识别地形图进行切分;对每个子图分别使用模型进行预测,集成子图的预测结果得到碎片化栅格地形图的比例尺.经过试验验证,本文方法的识别精度在97％左右,证明了本文方法的有效性.
Knowledge-guided intelligent recognition of the scale for fragmented raster topographic maps
Determining the topographic map scale is a critical basis for assessing the degree of confidentiality of topographic maps.In this study,we propose a solution to the challenge of estimating the scale of fragmented raster topographic maps by leveraging a priori knowledge of scale-related features,constructing an expert knowledge image pyramid dataset(EKIPD)un-der guided expert knowledge,and applying deep convolutional neural network algorithms to create a hybrid intelligent model that synergistically combines knowledge,data,and algorithm.The EKIPD dataset captures a representative sample distribu-tion of fragmented topographic maps of varying sizes,which enables us to statistically determine the optimal recognition size(ORS)for sub-map recognition.The ORS then serves as a stepping threshold to partition the topographic maps into recogniza-ble sub-maps.Each sub-map is independently processed through the model to obtain individual predictions,which are subse-quently integrated to infer the map scale.Experimental validation shows that this method achieves an accuracy of approximate-ly 97％,demonstrating its efficacy.

intelligentized surveying and mappingexpert knowledgehybrid intelligenceraster topographic mapscale rec-ognitiondeep convolutional neural network

任加新、刘万增、陈军、张蓝、陶远、朱秀丽、赵婷婷、李然、翟曦、王海清、周晓光、侯东阳、王勇

展开 >

中南大学地球科学与信息物理学院,湖南长沙 410083

国家基础地理信息中心,北京 100830

自然资源部时空信息与智能服务重点实验室,北京 100830

湖北珞珈实验室,湖北武汉 430079

中国矿业大学环境与测绘学院,江苏徐州 221116

中国测绘科学研究院,北京 100830

展开 >

智能化测绘 专家知识 混合智能 栅格地形图 比例尺识别 深度卷积神经网络

国家自然科学基金重大项目国家重点研发计划湖北珞珈实验室开放基金资助项目

423940622022YFB3904205220100037

2024

10.11947/j.AGCS.2024.20230005

测绘学报
中国测绘学会

测绘学报

CSTPCD北大核心
影响因子：1.602
ISSN：1001-1595
年,卷(期)：2024.53(1)
  • 29