Severe-occluded 3D object identification via region-based descriptions
作者：
Highlights：
•
• Regions provide reliable supports to separate objects from background.
• Regions benefit also descriptors based on depth and point clouds.
• Distributed encoding foster the model representation capability.
• Knowledge generalization from a very small set of short varied training instances.
• High object identification rates in severe occlusions and cluttered scenarios.
摘要
•Regions provide reliable supports to separate objects from background.•Regions benefit also descriptors based on depth and point clouds.•Distributed encoding foster the model representation capability.•Knowledge generalization from a very small set of short varied training instances.•High object identification rates in severe occlusions and cluttered scenarios.
论文关键词：3D-object identification,Severe-occlusion,Region-based PoI descriptions,Identification in clutter,Self-organized maps,Distributed encoding
论文评审过程：Received 27 December 2016, Revised 25 July 2017, Accepted 25 July 2017, Available online 3 August 2017, Version of Record 29 August 2017.
论文官网地址：https://doi.org/10.1016/j.image.2017.07.007
