[1]王学文,赵庆展,隆学丰,等.基于无人机多光谱和SOLO模型的防护林枯树提取方法[J].南京师范大学学报(工程技术版),2022,(01):046-51.[doi:10.3969/j.issn.1672-1292.2022.01.007]
 Wang Xuewen,Zhao Qingzhan,Long Xuefeng,et al.Extraction of Dead Trees from Shelter-Forest Based onUAV Multispectral and SOLO Model[J].Journal of Nanjing Normal University(Engineering and Technology),2022,(01):046-51.[doi:10.3969/j.issn.1672-1292.2022.01.007]
点击复制

基于无人机多光谱和SOLO模型的防护林枯树提取方法
分享到:

南京师范大学学报(工程技术版)[ISSN:1006-6977/CN:61-1281/TN]

卷:
期数:
2022年01期
页码:
046-51
栏目:
机器学习
出版日期:
2022-03-15

文章信息/Info

Title:
Extraction of Dead Trees from Shelter-Forest Based onUAV Multispectral and SOLO Model
文章编号:
1672-1292(2022)01-0046-06
作者:
王学文12赵庆展12隆学丰12胡 斌12
(1.石河子大学信息科学与技术学院,新疆 石河子 832003)(2.兵团空间信息工程技术研究中心,新疆 石河子 832003)
Author(s):
Wang Xuewen12Zhao Qingzhan12Long Xuefeng12Hu Bin12
(1.College of Information Science and Technology,Shihezi University,Shihezi 832003,China)(2.Geospatial Information Engineering Research Center,Xinjiang Production and Construction Corps,Shihezi 832003,China)
关键词:
无人机多光谱影像最佳指数因子样本生成SOLO防护林枯树提取
Keywords:
UAV multispectral imageoptimal exponential factorsample generationSOLO modeldead tree extraction of shelterbelt
分类号:
TP391
DOI:
10.3969/j.issn.1672-1292.2022.01.007
文献标志码:
A
摘要:
针对三北防护林树种混交比例低、灌水不足、空间分布不均匀、病虫害时发造成的防护林衰退问题,提出了一种小样本下基于无人机多光谱波段生成的SOLO(segmenting objects by locations)实例分割提取枯树的方法. 通过无人机搭载Micro MCA12 Snap多光谱相机获取高空间分辨率影像,将多光谱可见光波段(波段5、波段3、波段1)导出并标注样本,基于最佳指数因子选出排名前10的波段组合进行数据集的扩充,最后基于实例分割SOLO模型进行防护林枯树的提取. 实验结果表明,加入健康树样本,基于ResNet-101+FPN的SOLO模型AP从61.3%提升到63.8%,ResNet-50+FPN组合AP从60.7%提升到63.6%,同时进一步验证了这种样本增强方式的有效性.
Abstract:
Aiming at the decline of three-north shelter-forest caused by low mixed proportion of tree species,insufficient irrigation,uneven spatial distribution and frequent occurrence of diseases and pests,a method of extracting dead trees based on SOLO generated by UAV multi-spectral band under few-shot is proposed in this paper. The UAV is equipped with a Micro MCA12 Snap multispectral camera to obtain high spatial resolution images,derive and labele samples of multispectral visible bands(band 5,band 3,band 1). Based on the selected top 10 band combinations according to the best exponential factor,the dataset is expanded,and finally dead trees of shelter-forest are extracted based on the instance segmentation SOLO model. The experimental results show that with the addition of health tree samples,the SOLO model AP based on ResNet-101+FPN is increased from 61.3% to 63.8%,and the combined ResNet-50+FPN AP is increased from 60.7% to 63.6%. At the same time,the effectiveness of this sample enhancement method is further verified.

参考文献/References:

[1] WANG F,PAN X,GERLEIN S,et al. Vegetation restoration in Northern China:A contrasted picture[J]. Land Degradation & Development,2020,31(6):669-676.
[2]DENG C,ZHANG B,CHENG L,et al. Vegetation dynamics and their effects on surface water-energy balance over the Three-North Region of China[J]. Agricultural and Forest Meteorology,2019,275:79-90.
[3]COOPS N C M,JOHNSON M A,WULDER M A,et al. Assessment of quickbird high spatial resolution imagery to detect red attack damage due to mountain pine beetle infestation[J]. Remote Sensing of Environment,2006,103(1):67-80.
[4]SKAKUN R S,WULDER M A,FRANKLIN S E. Sensitivity of the thematic mapper enhanced wetness difference index to detect mountain pine beetle red-attack damage[J]. Remote Sensing of Environment,2003,86(4):433-443.
[5]陶欢,李存军,谢春春,等. 基于HSV阈值法的无人机影像变色松树识别[J]. 南京林业大学学报(自然科学版),2019,43(3):99-106.
[6]吕晓君,王君,喻卫国,等. 无人机监测林业有害生物初探[J]. 湖北林业科技,2016,45(4):30-33.
[7]HE K,GKIOXARI G,DOLLáR P,et al. Mask R-CNN[C]//Proceedings of the IEEE International Conference on Computer Vision. Venice,Italy,2017:2961-2969.
[8]任媛媛,张显峰,马永建,等. 基于卷积神经网络的无人机遥感影像农村建筑物目标检测[J]. 南京师范大学学报(工程技术版),2019,19(3):29-36.
[9]李越帅,郑宏伟,罗格平,等. 集成U-Net方法的无人机影像胡杨树冠提取和计数[J]. 遥感技术与应用,2019,34(5):939-949.
[10]王学文,赵庆展,韩峰,等. 机载多光谱影像语义分割模型在农田防护林提取中的应用[J]. 地球信息科学学报,2020,22(8):1702-1713.
[11]GAO X,SUN X,YAN M L,et al. 2018. Road extraction from remote sensing images by multiple feature pyramid network[C]//2018 IEEE International Geoscience and Remote Sensing Symposium. Valencia,Spain:IEEE:6907-6910.
[12]WANG Z C,DU L,MAO J S,et al. SAR targe detection based on SSD with data augmentation and transfer learning[J]. IEEE Geoscience and Remote Sensing Letters,2019,16(1):150-154.
[13]孙显,梁伟,刁文辉,等. 遥感边缘智能技术研究进展及挑战[J]. 中国图象图形学报,2020,25(9):1719-1738.
[14]POTLAPALLY A,MISHRA N,SAI P,et al. Instance segmentation in remote sensing imagery using deep convolutional neural networks[C]//2019 International Conference on Contemporary Computing and Informatics. Singapore:IEEE,2019.
[15]BOLYA D,ZHOU C,XIAO F,et al. YOLACT:Real-time instance segmentation[C]//Proceedings of the IEEE International Conference on Computer Vision. Seoul,South Korea,2019:9157-9166.
[16]WU T,HU Y,PENG L,et al. Improved anchor-free instance segmentation for building extraction from high-resolution remote sensing images[J]. Remote Sensing,2020,12:2910.
[17]WANG X,KONG T,SHEN C,et al. SOLO:segmenting objects by locations[C]//European Conference on Computer Vision. Glasgow:Springer,2020.
[18]邓颖,吴华瑞,朱华吉. 基于实例分割的柑橘花朵识别及花量统计[J]. 农业工程学报,2020,36(7):200-207.
[19]JOSé R G,BRAGRA V P,DALAGNOL R,et al. Tree crown delineation algorithm based on a convolutional neural network[J]. Remote Sensing,2020,12(8):1-27.
[20]JI W,WANG Z Y,WEN Q. A fusion method of hyperspectral image based on spectral high fidelity applied in spectrum retrieval of vegetation species[J]. Multispectral Hyperspectral and Ultraspectral Remote Sensing Technology Techniques and Applications,2014:9263.
[21]CHEN K,WANG J,PANG J,et al. MMDetection:open mmlab detection toolbox and benchmark[J]. arXiv Preprint arXiv:1906.07155,2019.
[22]PASZKE A,GROSS S,MASSA F,et al. Pytorch:an imperative style,high-performance deep learning library[C]//Proceedings of the Advances in Neural Information Processing Systems. Vancouver,Canada,2019.

备注/Memo

备注/Memo:
收稿日期:2021-08-31.
基金项目:国家重点研发计划项目(2017YFB0504203)、中央引导地方科技发展专项资金项目(201610011).
通讯作者:赵庆展,硕士,教授,研究方向:深度学习、机器学习、农业信息化、空间信息系统集成与服务研究. E-mail:zqz_inf@shzu.edu.cn
更新日期/Last Update: 2022-03-15