Petals' shape descriptor for blooming flowers recognition

This paper proposes a new descriptor to identify the petals' shape of a blooming flower based on the digital images captured in natural scene. The proposed descriptor can be used as one of features in computer aided flower recognition system beside the commonly used features such as number of petals and color. Experiments were conducted on the Malaysia flowers with same number of petals and with similar color across different species of flowers. 35 images from 7 species were used as the training set to set up the reference values of petals' shape descriptor and 7 new images were used as the testing set. The descriptor calculated from the testing set is then compared to the reference values from the training set to achieve the flowers recognition purpose. With the given set of data, complete success in full identification rate was obtained.