混合U-Net模型下南方丘陵地区耕地识别

doi:10.13474/j.cnki.11-2246.2023.0359

摘要/Abstract

摘要： 针对南方丘陵地区耕地零散、破碎、不规则导致的山垅田识别难度大、准确率低等问题,本文在U-Net网络基础上,引入高效通道注意力(ECA)和注意力门(AG)双注意力机制,提出了混合U-Net模型,并应用于福建省南安市2021年WorldView-2卫星影像的耕地提取。试验表明:混合U-Net模型识别的耕地取得了较好精度,总体精度达到93.42%,优于单一注意力机制模型ECA U-Net和U-Net模型,分别提升了9.75%和19%;混合U-Net模型在山区、半丘陵、平原3个不同地形的测试区域,耕地F1分数平均值分别为0.921 2、0.902 5、0.932 2,特别是在山区、半丘陵地区较ECA U-Net和 U-Net模型结果提升明显。在此基础上,对南安市耕地空间分布进行了分析,为山垅田撂荒复耕种粮、合理调整25°以上坡耕地、耕地数量管控提供了有效技术支持。

关键词: 耕地, 高分辨率影像, 注意力机制, U-Net, 南方丘陵地区

Abstract: In order to solve the difficulties and low accuracy of sloping cropland identification caused by scattered, fragmented and irregular cropland in southern hilly areas, based on the U-Net, the efficient channel attention(ECA) and the attention gate (AG) dual attention mechanism are introduced, and a hybrid U-Net model is proposed. This model is applied to extract cropland from WorldView-2 satellite images in Nan'an of Fujian province in 2021. Experiment shows that the hybrid U-shaped network model has achieved a good accuracy of 93.42%, which is better than a single-attention mechanism model (ECA U-Net and U-Net), and the accuracy has increased by 9.75% and 19% respectively. The average F1 scores of cropland in the hybrid U-Net model are 0.921 2, 0.902 5 and 0.932 2 respectively in the mountainous, semi hilly and plain test areas, especially in the mountainous and semi hilly areas. On this basis, the spatial distribution of cropland in Nan'an is analyzed, which provided effective technical support for abandoned hillside fields for grain cultivation, rational adjustment of slope cropland exceeding 25 degrees, and effective control of cropland quantity.

Key words: cropland, high-resolution image, attention mechanism, U-Net, southern hilly area

中图分类号:

吴瑞姣. 混合U-Net模型下南方丘陵地区耕地识别[J]. 测绘通报, 2023, 0(12): 57-62,111.

WU Ruijiao. Cropland recognition in southern hilly areas under a hybrid U-Net model[J]. Bulletin of Surveying and Mapping, 2023, 0(12): 57-62,111.

参考文献

[1] WHITE E V,ROY D P. A contemporary decennial examination of changing agricultural field sizes using Landsat time series data[J]. Geo: Geography and Environment,2015,2(1): 33-54.
[2] PATEL S K,MANI I,SRIVASTAVAET AP. The effect of tractor load and trafficking on wheat crop yield[EB/OL].2021-11-14[2023-03-88].https://www.researchgate.net/profile/Sanjay-Patel-6/publication/268218337_The_Effect_of_Tractor_Load_and_Trafficking_on_Wheat_Crop_Yield/links/55420aa20cf24107d392a199/The-Effect-of-Tractor-Load-and-Trafficking-on-Wheat-Crop-Yield.pdf.
[3] 蔡志文,何真,王文静,等. 基于多源国产高分卫星时空信息的米级分辨率耕地提取[J]. 遥感学报,2022,26(7): 1368-1382.
[4] 张海天,高懋芳,任超. 基于改进U-Net++的丘陵地区耕地地块深度分割与提取[J]. 航天返回与遥感,2022,43(4): 36-45.
[5] 周培诚,程塨,姚西文,等.高分辨率遥感影像解译中的机器学习范式[J].遥感学报,2021,25(1):182-197.
[6] 李国清,柏永青,杨轩,等.基于深度学习的高分辨率遥感影像土地覆盖自动分类方法[J]. 地球信息科学学报,2021,23(9): 1690-1704.
[7] 席一帆,孙乐乐,何立明,等. 基于改进Deeplab V3+网络的语义分割[J]. 计算机系统应用,2020,29(9): 178-183.
[8] SHELHAMER E,LONG J,DARRELL T. Fully convolutional networks for semantic segmentation[C]//Proceedings of 2016 IEEE Transactions on Pattern Analysis and Machine Intelligence.[S.l.]:IEEE,2016: 640-651.
[9] RONNEBERGER O,FISCHER P,BROX T. U-net: convolutional networks for biomedical image segmentation[M]//Lecture Notes in Computer Science. Cham: Springer International Publishing,2015: 234-241.
[10] ZHOU Zongwei,RAHMAN SIDDIQUEE M M,TAJBAKHSH N,et al.U-Net++:a nested U-net architecture for medical image segmentation[M]//Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support. Cham: Springer International Publishing,2018: 3-11.
[11] HE Kaiming,ZHANG Xiangyu,REN Shaoqing,et al. Deep residual learning for image recognition[C]//Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Las Vegas,NV:IEEE,2016: 770-778.
[12] 刘巍,吴志峰,骆剑承,等. 深度学习支持下的丘陵山区耕地高分辨率遥感信息分区分层提取方法[J]. 测绘学报,2021,50(1): 105-116.
[13] 陈玲玲,施政,廖凯涛,等. 基于卷积神经网络的高分遥感影像耕地提取研究[J]. 农业机械学报,2022,53(9): 168-177.
[14] WATKINS B,VAN NIEKERK A. Automating field boundary delineation with multi-temporal Sentinel-2 imagery[J]. Computers and Electronics in Agriculture,2019,167: 105078.
[15] WANG Qilong,WU Banggu,ZHU Pengfei,et al. ECA-net: efficient channel attention for deep convolutional neural networks[C]//Proceedings of 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Seattle,WA:IEEE,2020: 11531-11539.
[16] OKTAY O,SCHLEMPER J,LE FOLGOC L,et al. Attention U-net: learning where to look for the pancreas[EB/OL]. [2023-01-20].https://arxiv.org/abs/1804.03999.pdf.
[17] KINGMA D P,BA J. Adam: a method for stochastic optimization[EB/OL]. [2023-01-20].https://arxiv.org/abs/1412.6980.pdf.