[1]宋晓倩,张学艺,张春梅,等.基于深度迁移学习的酿酒葡萄种植信息提取[J].江苏农业学报,2020,(03):689-693.[doi:doi:10.3969/j.issn.1000-4440.2020.03.022]
 SONG Xiao-qian,ZHANG Xue-yi,ZHANG Chun-mei,et al.Extraction of wine grape planting information based on deep transfer learning[J].,2020,(03):689-693.[doi:doi:10.3969/j.issn.1000-4440.2020.03.022]
点击复制

基于深度迁移学习的酿酒葡萄种植信息提取()
分享到:

江苏农业学报[ISSN:1006-6977/CN:61-1281/TN]

卷:
期数:
2020年03期
页码:
689-693
栏目:
园艺
出版日期:
2020-06-30

文章信息/Info

Title:
Extraction of wine grape planting information based on deep transfer learning
作者:
宋晓倩1张学艺 2张春梅1李万春2
(1.北方民族大学计算机科学与工程学院,宁夏银川750021;2.中国气象局旱区特色农业气象灾害监测预警与风险管理重点实验室/宁夏气象防灾减灾重点实验室,宁夏银川750002)
Author(s):
SONG Xiao-qian1ZHANG Xue-yi2ZHANG Chun-mei1LI Wan-chun2
(1.School of Computer Science and Engineering, North Minzu University, Yinchuan 750021, China;2.Key Laboratory of Characteristic Agrometeorological Disaster Monitoring and Early Warning and Risk Management in Arid Regions, CMA/Key Laboratory for Meteorological Disaster Prevention and Reduction of Ningxia, Yinchuan 750002, China)
关键词:
酿酒葡萄信息提取高分数据深度学习迁移学习
Keywords:
wine grapesinformation extractionhigh resolution datadeep learningtransfer learning
分类号:
S127
DOI:
doi:10.3969/j.issn.1000-4440.2020.03.022
文献标志码:
A
摘要:
为利用遥感手段快速、精准提取宁夏贺兰山东麓酿酒葡萄种植信息,提出了一种基于深度迁移学习的酿酒葡萄种植信息提取方法。该方法以全卷积神经网络(Fully convolutional networks,FCN)为基础,利用高分二号卫星遥感资料,以地面采集样本数据进行网络模型训练,利用迁移学习方法将训练好的网络模型迁移到FCN网络模型中,对其进行初始化,避免过拟合问题的发生,其网络训练验证集准确率高达88.16%,较传统的基于深度学习方法准确率提高7.17个百分点。结果表明,基于深度迁移学习的贺兰山东麓酿酒葡萄种植信息提取检测准确率可达91.93%,检测召回率达到91.15%。
Abstract:
In order to extract the wine grape planting information of the eastern foot of Helan mountain by using remote sensing, a method based on deep transfer learning was proposed. On the basis of fully convolutional network(FCN), this method used the GF-2 remote sensing data to collect the sample data for the information extraction training. The trained network model was transferred to the FCN model by using the transfer learnin method. It was initialized in the network model to avoid over-fitting problem. The accuracy rate of its network training validation set was as high as 88.16%, which was 7.17 percentage points higher than that of traditional deep learning methods. The results showed that the accuracy rate of wine grape planting information extraction based on deep transfer learning at the eastern foot of Helan Mountain could reach 91.93%, and the recall rate reached 91.15%.

参考文献/References:

[1]张敏敏,代红军. 2015年贺兰山东麓产区赤霞珠葡萄成熟期品质的监测[J] . 农业科学研究,2016,37(2):29-33
[2]宋新欣. 宁夏贺兰山东麓葡萄产业品牌建设探讨[J]. 商业经济研究,2016(16):216-217.
[3]李卫国. 作物旱涝灾情遥感监测进展与思考[J] . 江苏农业学报,2013,29(6):1503-1506.
[4]王庆林,李卫国,王荣富. 基于不同时相遥感的冬小麦种植而积的提取[J]. 江苏农业学报,2014,30(6):1286-1291.
[5]徐新刚,李强子,周万村,等. 应用高分辨率遥感影像提取作物种植面积[J]. 遥感技术与应用,2008,23(1):17-23.
[6]曲景影,孙显,高鑫. 基于CNN模型的高分辨率遥感图像目标识别[J]. 国外电子测量技术,2016,35(8):45-50.
[7]张善文,谢泽奇,张晴晴. 卷积神经网络在黄瓜叶部病害识别中的应用[J]. 江苏农业学报,2018,34(1):56-61.
[8]孙云云,江朝晖,董伟,等. 基于卷积神经网络和小样本的茶树病害图像识别[J]. 江苏农业学报,2019,35(1):48-55.
[9]滕文秀,王妮,施慧慧,等. 结合面向对象和深度特征的高分影像树种分类[J]. 测绘通报,2019,505(4):38-42.
[10]黄云,唐林波,李振,等. 采用深度学习的遥感图像花生种植区域分类技术研究[J]. 信号处理,2019(4):617-622.
[11]KUSSUL N, LAVRENIUK M, SKAKUN S, et al. Deep learning classification of land cover and crop types using remote sensing data[J]. IEEE Geoscience & Remote Sensing Letters, 2017, 99:1-5.
[12]宫浩,张秀再,胡敬锋,等. 一种基于深度学习的遥感图像分类及农田识别方法[J]. 现代电子技术,2019,42(8):179-182.
[13]张学艺,李剑萍,秦其明,等. 几种干旱监测模型在宁夏的对比应用[J]. 农业工程学报,2009,25(8):18-23.
[14]潘佳颖. 贺兰山东麓葡萄基地两种类型土壤重金属分布特征与评价[D].银川:宁夏大学,2017.
[15]王竞,李磊,王锐,等. 不同生育期水分亏缺对酿酒葡萄光和特性、产量及品质的影响[J]. 江苏农业科学,2017,46(22):149-152.
[16]刘川,齐修东,藏文乾,等. 基于IHS变换的Gram-Schmidt改进融合算法研究[J]. 测绘工程,2018,27(11):9-14.
[17]姚新春,张亚亚. 基于EasyOrtho的高分二号卫星批量数据自动处理试验[J]. 北京测绘,2016(6):9-12.
[18]陈雨强. 图像与文本数据间的异构迁移学习[D]. 上海:上海交通大学,2012.
[19]张日升,朱桂斌,张燕琴. 基于卷积神经网络的卫星遥感图像区域识别[J]. 信息技术,2017(11):91-94.
[20]付辉敬,田铮. 遥感图像分割中的信息割算法[J]. 中国图象图形学报,2011,16(1):135-140.

备注/Memo

备注/Memo:
收稿日期:2019-10-08基金项目:国家自然科学基金项目(61461002);宁夏回族自治区青年拔尖人才培养工程项目(RQ0033);宁夏高等学校一流学科建设(电子科学与技术学科)项目(NXYLXK2017A07);宁夏回族自治区重点研发计划一般项目(2019BDE03011);研究生创新项目(YCX18059)作者简介:宋晓倩(1996-),女,山东日照人,硕士,研究方向为遥感图像处理。(Tel)15009611396;(E-mail)1040046117@qq.com通讯作者:张学艺,(Te
更新日期/Last Update: 2020-07-14