Preliminary study on a tree seedling grading system with machine vision

Abstract

The success of forest regeneration activities is strongly influenced by seedling quality. Better quality seedlings tend to have higher survival and growth rates. A PC-based machine vision system, which uses a Charge-Coupled-Device (CCD) camera and digital image processing techniques, was developed to increase the efficiency and accuracy of data collection in measuring morphological properties of forest seedlings in order to get better seedlings. The system relies on indirect illumination and specified hook, which is to hang seedling. Tests were conducted with a prototype of the vision system to measure Korean pine (Pinus koraiensis) seedlings. Machine vision measurements of shoot height, stem diameter, and root volume were compared with manual measurements. In the case of shoot height and stem diameter, the results showed high correlation (r = 0.993 and 0.929 calculated correlation coefficient, respectively) and high accuracy (6.14 and 0.34mm calculated standard error, respectively) between machine vision and manual measurements, but in the case of root volume factor, the correlation coefficient between the machine vision and manual measurement was not so good as 0.768. These results indicate the machine vision system is a useful tool for morphological measurement of tree seedlings and might have future applications in the automation of nursery grading.