Thai Blooming Flower Recognition Artidtaya Suppaiboonvong a Thesis Submitted in Partial Fulfillment of the Requirements for the Degree of Master of Science

The objective of this research is to build an automatic method for recognizing a Thai blooming flower based on minimum distance for finding similarities between each type of Thai blooming flower. The Euclidean distance technique was applied in order to improve performance of retrieval processes. The flower images were acquired from digital cameras which were taken at close range to ignore the background. The research focused on flowers that were already blooming for which the background was not considered, thus it was decided to take a photograph of the flowers at close range. The study used the characteristics of flowers to design the classification algorithms, which consist of eight-part features extraction from the whole flowers and three overlapping rectangles from the eight-part features to average the RGB colors. The experiments were conducted on more than 945 pictures from 23 species of Thai flowers. The training data set was at around 692 Thai blooming flower images. I tested the system by using 253 Thai blooming flower images for a training data set and 20 Thai blooming flower images for non-training data set. The non-training data set was conducted in order to gain a negative identification result, and therefore validate the overall system. The precision of the system was 99.12 % for the training set and 100 % for the non-training set.