Abstract—Modern description methods are used for plant classification through leaf recognition. These methods usually include color transformation, feature detection and description, dimension reduction, and classification. However, these methods use an original image as the input image from which to extract the features to be recognized. In this condition, computational complexity will increase. To reduce computational time, in the proposed method the Region of Interest (ROI) is extracted before extracting features from the image. Quality of image also plays an important role in increasing leaf classification rate. A good quality image gives better classification rate than noisy images. To extract features exactly from noisy images is very difficult which in-turn reduces leaf classification rate. To overcome problems occurring due to noisy image quality, background removal is done before extracting features from the image. That is, the proposed method includes color transformation, preprocessing (background removal and ROI extraction), feature extraction and description, and classification. In experiments and comparing results with and without preprocessing methods, the proposed method gives classification rate with an accuracy greater than 92.13% and the computational time in average is 133.94ms per leaf image.
Index Terms—leaf classification, background removal, ROI extraction, feature extraction, codebook creation
Cite: Yen-Ju Wu, Chun-Ming Tsai, and Frank Shih, "Improving Leaf Classification Rate via Background Removal and ROI Extraction," Journal of Image and Graphics, Vol. 4, No. 2, pp. 93-98, December 2016. doi: 10.18178/joig.4.2.93-98
Copyright © 2012-2021 Journal of Image and Graphics, All Rights Reserved