Exhaustive　investigations　of　the　ice　sheet　subsurface　can　be　carried　out　by　analyzing　the　information　contained　in　the　huge　archives　of　radar　grams　acquired　by　dedicated　radar　sounder　(RS)　instruments.　In　particularly,　an　automatic　segmentation　technique　enables　a　fast　and　objective　extraction　of　ice　subsurface　target　properties　on　wide　areas.　Here,　an　approach　which　automatically　segment　radar　image　at　the　pixel　level　using　capsule　Network　was　proposed.　Our　work　expands　the　use　of　capsule　networks　to　the　task　of　extraction　of　ice　subsurface　target　in　the　literature.　In　this　paper,　we　adopts　three　kinds　of　network　frameworks　for　the　task　of　extraction　of　ice　subsurface　target.　We　also　discuss　the　performance　of　squashing　function　on　the　segmentation　result.　Experimental　results　on　MCoRDS　datasets　confirm　the　performanceiveness　of　our　methods.　©　2019　Association　for　Computing　Machinery.