COLOR IMAGE SEGMENTATION AND LABELING THROUGH MULTIEDIT-CONDENSING

被引:15
作者
FERRI, F
VIDAL, E
机构
[1] UNIV POLITECN VALENCIA,DEPT SISTEMAS INFORMAT & COMPUTAC,CAMINO DE VERA S-N,E-46020 VALENCIA,SPAIN
[2] UNIV VALENCIA,DEPT INFORMAT & ELECTR,VALENCIA,SPAIN
关键词
NEAREST NEIGHBOR RULE; MULTIEDIT-CONDENSING; SCENE SEGMENTATION; PATTERN RECOGNITION; ROBOTICS;
D O I
10.1016/0167-8655(92)90091-D
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
A new method is proposed for detecting and locating objects of interest within a colour scene under verv strong variabilities in lighting conditions, object shape and pigmentation. The method is based on Nearest Neighbour classification and Multiedit-Condensing techniques and is applied to implement the vision subsystem of a robotic citric harvesting device. Experiments and results are reported showing the effectiveness of the method and illustrating its appropriateness to the proposed task.
引用
收藏
页码:561 / 568
页数:8
相关论文
共 12 条