Abstract
A new method of extracting various types of human knowledge from scenery images in linguistic form and labeling the various regions of unknown images in an effective manner is proposed. In this system, partitioning of the input/output space can be performed in a self-organizing manner. Rule extraction is performed in parallel and automatically by using a knowledge extraction network (KEN). Depending on the type of image to be learned, two-phase processing including a self-organization phase and an LMS learning phase is performed by several KENs. The overall network output is then unified and knowledge extraction and region recognition are performed linguistically. Software that performs a series of such operations has been implemented and it has been confirmed that knowledge edge extraction and better image recognition can be performed linguistically even if human knowledge is not provided.
| Original language | English |
|---|---|
| Pages (from-to) | 82-90 |
| Number of pages | 9 |
| Journal | Electronics and Communications in Japan, Part II: Electronics (English translation of Denshi Tsushin Gakkai Ronbunshi) |
| Volume | 86 |
| Issue number | 3 |
| DOIs | |
| Publication status | Published - 2003 Mar |
Keywords
- Fuzzy inference
- Image recognition
- Knowledge extraction
- Neural network
- Scenery image
ASJC Scopus subject areas
- Physics and Astronomy(all)
- Computer Networks and Communications
- Electrical and Electronic Engineering