{"id":77546,"date":"2024-04-15T15:27:06","date_gmt":"2024-04-15T06:27:06","guid":{"rendered":"https:\/\/www.waseda.jp\/inst\/research\/?p=77546"},"modified":"2024-08-21T11:16:20","modified_gmt":"2024-08-21T02:16:20","slug":"segmented-recurrent-transformer-with-cubed-3d-multiscanning-strategy-for-hyperspectral-image-classification%ef%bc%88published-in-political-behavior-april-2024%ef%bc%89","status":"publish","type":"post","link":"https:\/\/www.waseda.jp\/inst\/research\/news\/77546","title":{"rendered":"Segmented Recurrent Transformer with Cubed 3D-Multiscanning Strategy for Hyperspectral Image Classification\uff08Published in IEEE Transactions on Geoscience and Remote Sensing, April 2024\uff09"},"content":{"rendered":"<table class=\"table table-bordered table-colored-tbhd\" style=\"height: 550px; width: 100%; border-collapse: collapse; border-style: solid;\" border=\"1\">\n<tbody>\n<tr style=\"height: 78px;\">\n<td style=\"width: 19.0523%; height: 78px;\">Journal Title<br \/>\n\/\u63b2\u8f09\u30b8\u30e3\u30fc\u30ca\u30eb\u540d<\/td>\n<td style=\"width: 80.849%; height: 78px;\">IEEE Transactions on Geoscience and Remote Sensing<\/td>\n<\/tr>\n<tr style=\"height: 65px;\">\n<td style=\"width: 19.0523%; height: 80px;\">Publication Year and Month<br \/>\n\/\u63b2\u8f09\u5e74\u6708<\/td>\n<td style=\"width: 80.849%; height: 80px;\">April, 2024<\/td>\n<\/tr>\n<tr style=\"height: 55px;\">\n<td style=\"width: 19.0523%; height: 79px;\">Paper Title<br \/>\n\/\u8ad6\u6587\u30bf\u30a4\u30c8\u30eb<\/td>\n<td style=\"width: 80.849%; height: 79px;\">Segmented Recurrent Transformer with Cubed 3D-Multiscanning Strategy for Hyperspectral Image Classification<\/td>\n<\/tr>\n<tr style=\"height: 85px;\">\n<td style=\"width: 19.0523%; height: 85px;\">DOI<br \/>\n\/\u8ad6\u6587DOI<\/td>\n<td style=\"width: 80.849%; height: 85px;\"><a href=\"https:\/\/doi.org\/10.1109\/TGRS.2024.3384403\">10.1109\/TGRS.2024.3384403<\/a><\/td>\n<\/tr>\n<tr style=\"height: 59px;\">\n<td style=\"width: 19.0523%; height: 80px;\">\u00a0Author of Waseda University<br \/>\n\/\u672c\u5b66\u306e\u8457\u8005<\/td>\n<td style=\"width: 80.849%; height: 80px;\">ZHOU, Weilian(Assistant Professor, Faculty of Science and Engineering, Information, Production, and Systems Center):Lasr Author<\/td>\n<\/tr>\n<tr style=\"height: 68px;\">\n<td style=\"width: 19.0523%; height: 86px;\">Related Websites<br \/>\n\/\u95a2\u9023Web<\/td>\n<td style=\"width: 80.849%; height: 86px;\">&#8211;<\/td>\n<\/tr>\n<tr style=\"height: 138px;\">\n<td style=\"width: 19.0523%; height: 148px;\">Abstract<br \/>\n\/\u6284\u9332<\/td>\n<td style=\"width: 80.849%; height: 148px;\">This study introduces an innovative approach to hyperspectral imaging (HSI) classification by integrating convolution, recurrence, and self-attention mechanisms in a 3-D configuration. We address several challenges such as the 1) disruption of spectral continuity by traditional dimensionality reduction methods like PCA, 2) the overlooking of band-to-band continuous features in existing spatial-only 2-D multiscanning strategy, and 3) the limitations in model design by simply cascading recurrent neural networks (RNNs) with transformers for HSI analysis. Our solution involves three core components: 1) subband grouping with group-wise convolution for refined dimension reduction; 2) a novel cubed 3-D-multiscanning technique enabling thorough multidirectional analysis in both spectral and spatial domains; and 3) the development of a Cubic-Net framework with a specially designed segmented recurrent transformer (SRT). This SRT is tailored to effectively utilize spectral continuity along with spatial contextual features, overcoming common sequential data analysis challenges seen in RNNs and transformers. Furthermore, our feature fusion strategy successively integrates \u201cshort-term\u201d and \u201clong-term\u201d SRT features, thereby enhancing the model\u2019s ability to process both spectral and spatial features effectively. Experimental results from three public HSI datasets indicate our method\u2019s improved performance over existing baselines and state-of-the-art methods. This research offers a new perspective on 3-D sequential HSI classification.<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n","protected":false},"excerpt":{"rendered":"<p>Journal Title \/\u63b2\u8f09\u30b8\u30e3\u30fc\u30ca\u30eb\u540d IEEE Transactions on Geoscience and Remote Sensing Publication Year and Month \/\u63b2\u8f09\u5e74\u6708 Ap [&hellip;]<\/p>\n","protected":false},"author":3,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[],"tags":[218,217],"class_list":["post-77546","post","type-post","status-publish","format-standard","hentry","tag-impact-en","tag-impact"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.waseda.jp\/inst\/research\/wp-json\/wp\/v2\/posts\/77546","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.waseda.jp\/inst\/research\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.waseda.jp\/inst\/research\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.waseda.jp\/inst\/research\/wp-json\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/www.waseda.jp\/inst\/research\/wp-json\/wp\/v2\/comments?post=77546"}],"version-history":[{"count":2,"href":"https:\/\/www.waseda.jp\/inst\/research\/wp-json\/wp\/v2\/posts\/77546\/revisions"}],"predecessor-version":[{"id":78263,"href":"https:\/\/www.waseda.jp\/inst\/research\/wp-json\/wp\/v2\/posts\/77546\/revisions\/78263"}],"wp:attachment":[{"href":"https:\/\/www.waseda.jp\/inst\/research\/wp-json\/wp\/v2\/media?parent=77546"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.waseda.jp\/inst\/research\/wp-json\/wp\/v2\/categories?post=77546"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.waseda.jp\/inst\/research\/wp-json\/wp\/v2\/tags?post=77546"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}