Authors:
Yuli Chen
1
;
Yide Ma
2
;
Dong Hwan Kim
3
and
Sung-Kee Park
3
Affiliations:
1
Lanzhou University and Korea Institute of Science and Technology, China
;
2
Lanzhou University, China
;
3
Korea Institute of Science and Technology, Korea, Republic of
Keyword(s):
Simplified Pulse Coupled Neural Network (SPCNN), Image Segmentation, Object Recognition, Region-based Matching.
Related
Ontology
Subjects/Areas/Topics:
Image Processing
;
Informatics in Control, Automation and Robotics
;
Intelligent Control Systems and Optimization
;
Neural Networks Based Control Systems
;
Robotics and Automation
;
Vision, Recognition and Reconstruction
Abstract:
The aim of the paper is to propose a region-based object recognition method to identify objects from complex real-world scenes. The proposed method firstly performs a colour image segmentation by a simplified pulse coupled neural network (SPCNN) model, and the parameters of the SPCNN are automatically set by our previously proposed parameter setting method. Subsequently, the proposed method performs a region-based matching between a model object image and a test image. A large number of object recognition experiments have proved that the proposed method is robust against the variations in translation, rotation, scale and illumination, even under partial occlusion and highly clutter backgrounds. Also it shows a good performance in identifying less-textured objects, which significantly outperforms most feature-based methods.