A high-resolution feature network image-level classification method for hyper-spectral image
Hyperspectral image(HSI)classification methods based on deep learning usually slice hyperspectral images into lo-cal-patches as the input of the model,which not only limits the acquisition of long-distance space-spectral information associa-tion,but also brings a lot of extra computational overhead.The image-level classification method with global image as input can effectively avoid these defects.However,the detail loss during information recovery of the existing image-level classifica-tion methods based on feature serial flow pattern of fully convolutional network(FCN)will lead to problems such as low classi-fication accuracy and poor visual effect of the classification map.Therefore,this paper proposes a high-resolution feature net-work(HRNet)image-level classification method for hyperspectral image,which performs parallel computation and cross fu-sion of multi-resolution features of images while maintaining high-resolution features throughout the whole process,thus allevi-ating the information loss caused by the traditional serial flow pattern of features.Simultaneously,we propose a jointly-su-pervised training strategy of multi-resolution feature and a vote classification strategy,so as to further improve the classification per-formance of the model.Four public hyperspectral image datasets are used to verify the proposed method.Experimental results show that compared with the existing advanced classification methods,the proposed method can obtain competitive classification results,sig-nificantly reduce the training and classification time at the same time,and is more time-sensitive in practical application.In order to as-sure the reproducibility of method,we will open the code at https://github.com/sssssyf/fast-image-level-vote.
万方数据
维普
