Research on forest information extraction based on improved dilated convolutional UNet network
To improve the intelligence and efficiency of large-scale forest information extraction,a residual connected dilated convolu-tional UNet(RCD-UNet)network has been proposed.The residual connected dual convolutional module is organically combined with the traditional UNet network to improve model performance.An experiment on forest extraction in Nanjing has been conducted using high-resolution satellite remote sensing images of Gaofen-2 as the data source.The results indicate that the proposed method can en-hance the model's ability to perceive the context by introducing an atrous spatial pyramid pooling module.The overall accuracy of forest extraction is 95.44％,and the Kappa coefficient is 82.48％,which meets the requirements for efficient and accurate extraction of forest resource spatial structure information,providing technical support for forest resource management and investigation.
GF-2dilated convolution UNetresidual connectionforest information extraction
万方数据
维普
