Shape Grading of Pimento Using Image Processing

• Main Authors: Hung-Tzu Su, 蘇鴻賜
• Other Authors: Hsien-Huang Wu
• Format: Others
• Language: zh-TW
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/37949105077406706156

碩士 === 國立雲林科技大學 === 電機工程系碩士班 === 94 === Grading the agricultural product is quite an important work in the transportation and selling process. The uniform and high quality product is the essential condition to enhance good marketing image and to get high selling prices. Two major studies were accomplished and successfully picked out the high quality pimentos. The first study was to examine the color the mature pimentos by RGB and HSV color model fixed scope method. The mature pimento''s color was defined as it meat the specific scope. The following study establish the was to basic rectangle primarily shape, from the manual graded pimento''s images, by its feature of length, contour length, width, angles…etc(for ranking the grading criterion). We also obtained the cluster centers, mean vectors and covariance matrix from the training samples by using K-means and Quadratic classifier. In the experimental classifying process, K-means algorithm performed the identification of the similarity of the test sample''s feature values and the values cluster centers. The Quadratic classifier made the classifications based on Gauss probability density function. The experiments showed satisfying results that RGB and HSV color model reliability are 93% and 97%. For grading reliability, K-means and Quadratic are 92% and 95% respectively.