Non-destructive growth measurement of selected vegetable seedlings using orthogonal images
Journal
Transactions of the American Society of Agricultural Engineers
Journal Volume
48
Journal Issue
5
Start Page
1953
End Page
1961
ISSN
0001-2351
2151-0059
Date Issued
2005
Author(s)
Chien, C.F.
Abstract
A non-destructive measurement method of plant features using image processing technique provides a means to analyze the continuous growth process of a plant. In a previous study, we developed a non-destructive measurement method by using elliptical Hough transform to search for seedling leaves from a top-view image. To improve the accuracy of the leaf number estimation and leaf area measurement, this study further incorporates two side-view images with a top-view image to extract and reconstruct the three-dimensional structure of selected vegetable seedlings, and thus to measure the features related to plant growth. By registering correspondences between the three orthogonal images, the position, orientation, and dimensions of leaves in a seedling can be more exactly derived than from a single top-view image alone. The leaf skeletons traced in the side-view images can be used to correct the projected leaf area in the top-view image for better estimation of leaf area. Experiments were performed to test the performance of the measurement system by comparing the non-destructively estimated leaf area and the actual leaf area of selected vegetable seedlings. The average relative errors of total leaf estimation for cabbage and broccoli seedlings were 14.5% and 13.1%, respectively, using a single top-view image. By incorporating orthogonal images, the relative errors were significantly reduced to 1.6% and 4.9%, respectively. Using this non-destructive measurement system, the continuous growth of vegetable seedlings can be effectively measured.
Subjects
Cluster analysis
Hough transform
Image processing
Orthogonal images
Plant growth
SDGs
Publisher
American Society of Agricultural and Biological Engineers (ASABE)
Type
journal article
