Photovoltaic roof extraction from material and angle attribute filtered method

doi:10.13474/j.cnki.11-2246.2021.0085

Abstract

Abstract: Aiming at the high-resolution satellite image photovoltaic roof extraction problem under the background of sample imbalance caused by various types of urban photovoltaic roof, a material and angle attribute filtered method is proposed. By selecting satellite images of Nankai district, Hongqiao district, Heping district, and Hedong district in Tianjin, the sample balanced data set is acquired using the two filtered material and angle attributes labeled on the photovoltaic roof. This paper finally gets the photovoltaic roof extraction result by properly setting parameters of the U-Net algorithm integrated under the TensorFlow framework. As a comparison, the photovoltaic roof is extracted without using two filtered material and angle attributes at the same time. The experiment shows that the method proposed in this paper has higher extraction accuracy, and especially for the uncommon photovoltaic roof in the study area, it can extract more complete and accurate result with clear boundary.

Key words: photovoltaic roof, sample imbalance, material, angle, U-Net algorithm

CLC Number:

WANG Shouzhi, ZHANG Fukun, ZHU Pengfei, ZHAN Hao, ZHANG Yunjiao, XI Ge. Photovoltaic roof extraction from material and angle attribute filtered method[J]. Bulletin of Surveying and Mapping, 2021, 0(3): 96-99,151.

References

[1] 吴扬昀, 魏立明, 柴梓淳. 光伏发电系统在建筑节能中的应用研究探析[J]. 北方建筑, 2019, 4(4):56-59.
[2] 杨扬, 武文佳. 基于无人机影像的屋顶检测与光伏自动排布方法[J]. 现代制造技术与装备, 2019(6):86-88.
[3] FARABET C, COUPRIE C, NAJMAN L, et al. Learning hierarchical features for scene labeling[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013, 35(8):1915-1929.
[4] CHARALAMPOUS K, KOSTAVELIS I, AMANATIADIS A, et al. Sparse deep-learning algorithm for recognition and categorisation[J]. Electronics Letters, 2012, 48(20):1265-1266.
[5] OLSHAUSEN B A, FIELD D J. Emergence of simple-cell receptive field properties by learning a sparse code for natural images[J]. Nature, 1996, 381(13):607-609.
[6] 金永涛, 杨秀峰, 高涛, 等. 基于面向对象与深度学习的典型地物提取[J]. 国土资源遥感, 2018, 30(1):22-29.
[7] 徐金晓, 方圆. 深度学习在高分辨率遥感影像解译中的应用研究[J]. 通信电源技术, 2019, 36(8):68-69.
[8] 许慧敏. 基于深度学习的U-Net模型的高分辨率遥感影像分类方法研究[D]. 成都:西南交通大学, 2018.
[9] 许玥. 基于改进Unet的遥感影像语义分割在地表水体变迁中的应用[D]. 重庆:重庆师范大学, 2019.
[10] 张涛. 基于深度学习的遥感图像目标分类的研究[D]. 太原:中北大学, 2018.
[11] 王守志, 邢立新, 仲波, 等. 基于Landsat-8 OLI和GF-1 PMS数据融合的铁染蚀变信息提取[J]. 遥感与技术应用, 2016, 31(5):950-957.
[12] 张蕊. 基于激光点云的复杂三维场景多态目标语义分割技术研究[D]. 郑州:战略支援部队信息工程大学, 2018.
[13] 敖焕轩. 基于深度学习与单目视觉SLAM的三维语义场景建模研究[D]. 南京:东南大学, 2018.
[14] 刘文涛. 基于卷积神经网络的城市区域建筑物自动提取研究[D]. 成都:电子科技大学, 2018.
[15] 陈斌, 王宏志, 徐新良, 等. 深度学习GoogleNet模型支持下的中分辨率遥感影像自动分类[J]. 测绘通报, 2019(6):29-33.