Preliminary design of a recognition system for infected fish species using computer vision

Abstract

For the purpose of classification of fish species, a recognition system was preliminary designed using computer vision. In the first place, pictures were pre-processed by developed programs, dividing into rectangle pieces. Secondly, color and texture features are extracted for those selected texture rectangle fish skin images. Finally, all the images were classified by multi-class classifier named SVMs. The experiment showed that color and texture are the appropriate features for fish species classification. The multi-class classifier based on SVM will be developed for further work. © 2012 IFIP International Federation for Information Processing.