摘要
土地卫片执法工作存在时间紧、任务重,难以全部审核等工作难点。本文基于深度学习技术,进行了举证照片智能识别技术在土地卫片执法审核工作中的应用探索。结合实际需求,本文选取DenseNet作为图片预测的主要模型结构,DenseNet在保证特征传递的同时有效减少了参数量,做到效率与精度兼顾。经与人工审核意见对比验证,结果显示,从高标准“通过”要求角度出发实现的智能识别辅助审核“通过”结果正确率较高,说明该方法用于土地卫片执法工作中进行辅助审核是可行的,可大幅提高内业审核比例,并为人工审核缩小工作范围。
Problems exist in the land law enforcement inspection using satellite images,such as time constraints,heavy tasks,and difficulties in complete inspection.Based on deep learning,this paper explores the application of intelligent recognition technology on evidence photo examination in land law enforcement inspection using satellite images.Combining with the actual requirements,this paper selects DenseNet as the main model structure of image prediction.DenseNet not only ensures feature transmission,but also effectively reduces the number of parameters,achieving both efficiency and accuracy.Compared with the manual examining opinion,the results show that the accuracy of intelligent recognition computer-assisted examining"pass"experimental data based on the high"pass"standard is superior,which indicates that this method is feasible for computer-assisted examining evidence photos,and can greatly increase the proportion of indoor examining,and reduce the scope of work for manual examining.
作者
刘剑
杨志刚
龙丽红
王腾
张晶
常圆
LIU Jian;YANG Zhigang;LONG Lihong;WANG Teng;ZHANG Jing;CHANG Yuan(Surveying and Mapping Institute of Land and Resources of Guangdong Province,Guangzhou 510500,China;Shantou University,Shantou 515063,China)
出处
《测绘与空间地理信息》
2021年第9期86-89,94,共5页
Geomatics & Spatial Information Technology
关键词
深度学习
智能识别
土地卫片执法
举证照片
辅助审核
deep learning
intelligent recognition technology
land law enforcement inspection using satellite images
evidence photos
computer-assisted inspection
