Semantic Segmentation Of Winter Wheat In High-resolution Remote Sensing Images Based On Improved U-Net Model

Accurate acquisition of winter wheat planting distribution information is of great significance for agricultural management and food security.In recent years,with the development of remote sensing satellite technology and artificial intelligence,semantic segmentation of agricultural remote sensing images using convolutional neural networks has become one of the current research hotspots.Although the existing convolutional neural network makes the semantic segmentation of agricultural remote sensing images intelligent,due to the rich texture features,spatial information and complex geometric shapes of high-resolution agricultural remote sensing images,agricultural images are prone to poor feature representation and insufficient utilization of spatial information during semantic segmentation,resulting in rough prediction results.Aiming at the above problems,this paper proposes a high-resolution remote sensing image semantic segmentation model CBAM-ASPP-U-Net based on mixed attention mechanism CBAM and empty space pyramid pooling ASPP.The model uses a codec structure and an end-to-end approach to achieve semantic segmentation of high-resolution agricultural remote sensing images.There are two main innovations :(1)In order to solve the problem of poor feature representation ability in semantic segmentation of agricultural remote sensing images,CBAM-ASPP-U-Net semantic segmentation module adds a hybrid attention mechanism module CBAM on the basis of U-Net network structure,focusing on feature-intensive spatial regions and channels that contribute greatly to semantic segmentation results.(2)In order to solve the problem of insufficient utilization of spatial information in semantic segmentation of agricultural remote sensing images,the CBAM-ASPP-U-Net semantic segmentation model adds the hollow space pyramid pooling module ASPP on the basis of the traditional codec structure,and uses the multi-level sampling rate of the hollow convolution parallel to broaden the receptive field of the network,so that the semantic information obtained by the network is more abundant,so as to achieve the purpose of global context semantic information coding.In order to verify the effectiveness of the CBAM-ASPP-U-Net semantic segmentation model proposed in this paper,the semantic segmentation of winter wheat is compared on the Sentinel-2A remote sensing image data set of Qixian County,Kaifeng.The experimental results prove the effectiveness of the improved semantic segmentation model,which has stronger feature representation ability and can obtain more accurate semantic segmentation results of winter wheat.