Oncidium Cut Flower Grading with Machine Vision

• Main Author: 席友亮
• Other Authors: Fang-Fan Lee
• Format: Others
• Language: zh-TW
• Published: 2001
• Online Access: http://ndltd.ncl.edu.tw/handle/16345312459828338339

碩士 === 國立中興大學 === 農業機械工程學系 === 89 === The objective of this thesis is to use digital image processing techniques to extract feature parameters of oncidium cut flowers for grading. A human-machine interface was also developed for the future grading machine. In this study, two color images, namely the flower image and the stem image, were grabbed for each cut flower. The flower image was used to determine the length of the flower part and the stem image was used to determine the length of the stem part. The square-frame method was utilized to find the staring point of the flower part. The distance between the starting point and the end of the flower part was the length of the flower part. The distance between the starting point of the flower part and the end of the stem part was the length of the stem part. A rotation-tracking method developed by the author was used to find the edge points of the image. The method of least squares was employed to determine the similar line of the stem. Then, a 42mm width region was deleted from both sides of the similar line and the images of the branches were disconnected. These isolated branches were considered as the blobs of the processed image of the cut flower. The area and length of the blob, the stem’s area of the blob, and the length of the cut line of the blob were input to a neural network to determine the number of branches in each blob. The lengths of the flower part and stem part, and the number of the branches were used to grade the cut flowers according to the grading criteria. In addition, the projected area and boundary length of the flower part, the lengths of the flower part and stem part, the stem diameters of the cut flower in the middle and at the end of the stem were input to the neural network to grade the cut flowers. Using the grading criteria to grade 150 cut flowers, a 72% grading accuracy rate was obtained. Using the artificial neural network to grade 97 cut flowers, a 79% grading accuracy rate was obtained.