Herb flower recognition system (HFRS)

Abstract

The objective of this research is to build an automatic method for recognizing a Thai herb flower based on the Minimum Distance Method. The herb flower images, acquired from a digital camera, are taken in the real environment. We use the characteristics of herb flowers to design our classification algorithms, which consist of the average red, green and blue (RGB) colors, the herb flower size and the edge of petals feature. The experiments are conducted on more than 380 pictures from 16 species of herb flowers. The training data set is around 220 pictures. We test the system by using 110 pictures for a training data set and 50 pictures for an un-training data set. The precision rates of the recognition system are 98.18 percent and 94 percent, respectively. The average access time is 0.87 seconds per image. © 2010 IEEE.