Abstract
The recent emergence of high-resolution Synthetic Aperture Radar (SAR) images leads to massive amounts of data. In order to segment these big remotely sensed data in an acceptable time frame, more and more segmentation algorithms based on deep learning attempt to take superpixels as processing units. However, the over-segmented images become non-Euclidean structure data that traditional deep Convolutional Neural Networks (CNN) cannot directly process. Here, we propose a novel Attention Graph Convolution Network (AGCN) to perform superpixel-wise segmentation in big SAR imagery data. AGCN consists of an attention mechanism layer and Graph Convolution Networks (GCN). GCN can operate on graph-structure data by generalizing convolutions to the graph domain and have been successfully applied in tasks such as node classification. The attention mechanism layer is introduced to guide the graph convolution layers to focus on the most relevant nodes in order to make decisions by specifying different coefficients to different nodes in a neighbourhood. The attention layer is located before the convolution layers, and noisy information from the neighbouring nodes has less negative influence on the attention coefficients. Quantified experiments on two airborne SAR image datasets prove that the proposed method outperforms the other state-of-the-art segmentation approaches. Its computation time is also far less than the current mainstream pixel-level semantic segmentation networks.
Highlights
Synthetic Aperture Radar (SAR) image segmentation is a foundation for many high-level interpretation tasks
In order to improve the computational efficiency of the segmentation algorithms in the test phase, we propose to use the Graph Convolution Network (GCN) to segment the nodes of graph structure data
The superpixels belonging to urban, farmland, and river basically had similar numbers in boTthabthlee1trgaiivneinsgthaendnutemstbienrgs soeftsth. eThsuepqeurapnixtietlys ooff bfiavcekgcaroteugnodriseuspinertphiexetlrsaiwniansgmauncdhtleasrtginegr tsheatsn. tAhsatcaonf tbheeseoethne, rthfeousurpceartpegixoerlisebs,ewlohnigcihngcotonfuorrbmans,tfoatrhmelapnrdac, taincadlrciovnerdbitaiosinc.alIlnymhaodstsiomf itlhaer npuramctbicearsl ainppbloicthattihone,twraeinwinegreaonndlyteisntitnergessteetds. iTnhaesqmuaalnl tpitayrtooffblaacnkdgrsopuecnidessiunptehrepilxareglsewscaesnme.uch larger than that of the other four categories, which conforms to the practical condition
Summary
SAR image segmentation is a foundation for many high-level interpretation tasks. Accurate segmentation can greatly reduce the difficulty of subsequent advanced tasks (e.g., target recognition [1,2], target detection [3], change detection [4], etc.). In order to improve the computational efficiency of the segmentation algorithms in the test phase, we propose to use the Graph Convolution Network (GCN) to segment the nodes of graph structure data. Non-spatial methods aim at transforming graph-structure data into the Euclidean structure by redefining the neighbour regions of nodes [27], so that all the nodes have the same number of the adjacent nodes and the traditional CNN [28] can process the data These types of approaches generally have two steps: (1) select the most representative nodes to form the sequence of the nodes to be segmented, and (2) define a fixed size neighbouring field for each selected node. A novel attention GCN is proposed for node segmentation on the graph-structure data.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
