Vegetation extraction in coastal zones based on GF-2
For the existing extraction methods for vegetation in coastal zones,there are problems such as a long investigation period and great difficulty in the investigation,and it is difficult to meet the need of rapidly obtaining the large-scale vegetation status in the coastal zones.To address these issues,this paper studied the coastal area of Dalian City,Liaoning Province and Gaofen-2(GF-2)multispectral images and proposed an extraction method for vegetation in coastal zones based on improved U-shaped convolutional neural networks(Res_UNet).Firstly,the method produced an extraction dataset for vegetation in coastal zones based on sub-meter high-resolution images.Secondly,it fused the residual network(ResNet18)with a network depth of 18 layers and the U-Net network and introduced the residual module into U-Net to improve the fitting accuracy of the model.Finally,performance evaluation experiments were conducted on a test set consisting of 1 200 independent test samples.The results show that the Res_UNet model improves the average intersection over union(IOU)I and F1 by 2.554%and 1.949%compared with the U-Net model.The method proposed in this paper can realize the rapid extraction of large-scale vegetation in coastal zones and provide support for the construction of ecological civilization in coastal zones and the dynamic statistics of natural shoreline retention rate.
high-resolution imagevegetation in coastal zonenatural shoreline retention ratedeep learningvegetation extraction
万方数据
维普
