A vision-based hybrid approach for identification of Anthurium flower cultivars

Abstract

A hybrid approach was developed for highly accurate and effective identification of Anthurium flower cultivars in a computer vision-based sorting machine. Anthurium flowers have a small spike-shaped inflorescence called spadix. These flowers are distinguishable according to the color scheme of the spadix region. In the developed cultivar classification algorithm, the spadix region of test images was detected using the Viola-Jones object detection algorithm. The Viola-Jones detector was trained by positive images prepared from different cultivars of Anthurium flower, and the Oxford Flowers 17 dataset was used as negative images. Then, the detected region as Region of Interest (ROI) matched with images of various cultivars at different sizes and angles of rotation templates as a multi-template matching approach, in which each image was representative of a specified cultivar. The experiment results indicate that the proposed technique has acceptable performance in detecting the spadix region and inspiring performance in classifying the flower cultivars. At different conditions of the templates used for classification, the computation time as a critical criterion for real-time classification was less than 0.5 s, with the classification accuracy of more than 99%. In an automatic grading machine for flowers, cultivar classification of flowers is an important step for subsequent grading tasks.