U-Net模型对不同空间分辨率防护林提取精度的影响

Influence of U-Net model on the accuracy of shelter forest extraction with different spatial resolutions

针对当前无人机影像获取精度高但数据规模小的问题,本文提出通过U-Net模型探讨不同空间分辨率对防护林提取精度的影响。以CW-20复合翼无人机搭载Micro MAC12 Snap多光谱传感器获取的300 m(空间分辨率0.15 m)、400 m(空间分辨率0.20 m)、500 m(空间分辨率0.25 m)3种不同高度的遥感影像为例,试验结果证明,3种不同高度的影像提取精度误差在1.3%以内;MIoU误差在3.7%以内。空间分辨率对防护林提取精度的影响较小,高空间分辨率的影像数据并不能显著提升防护林的提取精度。本文为大规模农林业遥感监测的数据源获取提供了理论依据。

Aiming at the problem of high-accuracy of UAV image acquisition but small data scale,this paper proposes to use U-Net model to explore the influence of different spatial resolution on the extraction accuracy of farmland shelterbelts.Take the CW-20 compound-wing UAV equipped with Micro MAC12 Snap multispectral sensor obtained three different height of 300 m(spatial resolution 0.15 m),400 m(spatial resolution 0.20 m)and 500 m(spatial resolution 0.25 m)as an example.For remote sensing images,experiments results have shown that the accuracy error of image extraction at three different height is within 1.3%,and the accuracy error of MIoU is within 3.7%.The spatial resolution had little effect on the extraction accuracy of shelterbelts.The image data with high spatial resolution can not significantly improve the extraction accuracy of shelterbelts.This study provides a theoretical basis for the acquisition of large-scale agricultural and forestry remote sensing monitoring data sources.