1Graduate School of Engineering, Tokyo University of Agriculture and Technology, Tokyo, Japan
2Graduate School of Life Science and Systems Engineering, Kyushu Institute of Technology, Kitakyushu, Japan
Journal of Computer Sciences and Applications.
2015,
Vol. 3 No. 1, 1-10
DOI: 10.12691/jcsa-3-1-1
Copyright © 2015 Science and Education PublishingCite this paper: Satoru Yokota, Jiang Li, Yuichi Ogishima, Hiromasa Kubo, Hakaru Tamukoh, Masatoshi Sekine. Self-Learning of Feature Regions for Image Recognition.
Journal of Computer Sciences and Applications. 2015; 3(1):1-10. doi: 10.12691/jcsa-3-1-1.
Correspondence to: Hakaru Tamukoh, Graduate School of Life Science and Systems Engineering, Kyushu Institute of Technology, Kitakyushu, Japan. Email:
yokota@sekine-lab.ei.tuat.ac.jp, tamukoh@brain.kyutech.ac.jpAbstract
Mobile systems are used in various environments. Thus, it is practical for image recognition systems to autonomously learn template images that are adaptive to objects in their various environments. However, learning the features of such objects requires large-scale computation and complex control. Hence, we propose an image recognition system that selects and learns regions that have a given object's features. This system is designed as a hardware/software (hw/sw) complex system with the multi-dimensional field programmable gate array (FPGA) “Vocalise.” This study discusses the possibility of dynamically building image databases and of real-time learning using the proposed image recognition system. Results indicate that the learning speed of the proposed method is estimated to be 1.4 × 103 faster than that obtained with a conventional software method. This suggests the possibility of real-time learning.
Keywords