首页> 外文会议>International Conference on Computer Vision Systems >Architecture for Image Labelling in Real Conditions
【24h】

Architecture for Image Labelling in Real Conditions

机译:实际条件下图像标签的架构

获取原文

摘要

A general model for the segmentation and labelling of acquired images in real conditions is proposed. These images could be obtained in adverse environmental conditions, such as faulty illumination, non-homogeneous scale, etc. The system is based on surface identification of the objects in the scene using a database. This database stores features from series of each surface perceived with successive optical parameter values: the collection of each surface perceived at successive distances, and at successive illumination intensities, etc. We propose the use of non-specific descriptors, such as brightness histograms, which could be systematically used in a wide range of real situations and the simplification of database queries by obtaining context information. Self-organizing maps have been used as a basis for the architecture, in several phases of the process. Finally, we show an application of the architecture for labelling scenes obtained in different illumination conditions and an example of a deficiently illuminated outdoor scene.
机译:提出了一种在实际条件下分割和标记的分割和标记的一般模型。这些图像可以在不利的环境条件下获得,例如错误的照明,非均匀尺度等。该系统基于使用数据库的场景中对象的表面识别。该数据库存储来自连续光学参数值的每个表面的系列的特征:在连续距离处感知的每个表面的集合,并且在连续的照明强度等中。我们提出了使用非特定描述符,例如亮度直方图,这可以通过获取上下文信息来系统地用于各种实际情况和简化数据库查询。在过程的几个阶段,自组织地图已被用作架构的基础。最后,我们展示了用于在不同照明条件中获得的标记场景的架构的应用以及不可忽视的户外场景的示例。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号