Articles
DETERMINING SEEDLING CHARACTERISTICS USING COMPUTER VISION AND ITS APPLICATION TO AN EXPERT SYSTEM FOR GRADING SEEDLINGS
The camera was mounted approximately 1.2 m above the lettuce seedlings.
Partitioning partially overlapped leaves into single leaves and tracing the edge of each leaf was accomplished using the contraction and dilatation algorithms, and finally "AND" operation with the original binary image.
Subsequently the morphological characteristics such as number of leaves, area, length and width of individual leaves were determined.
To determine the distribution and expansion features of leaves, the median lines of individual leaves were extracted by the thinning algorithm.
These characteristics will be helpful to the analysis of seedling growth and development as well.
The expert system was established using commercially available hardware (AI Station) and software (XLAI) which were specifically designed for the applications of artificial intelligence.
The knowledge base was developed using a typical form of production rules which are expressed as "IF – THEN -" with certainty factor ranging from -1 to +1. Backward-chaining was used to reason.
The expert system was capable of grading a block of 64 lettuce seedlings in approximately one minute based on the extracted characteristics from the image analysis, the environmental condition and the seedling age.
Seedlings are graded into three quality ranks; good, fair and poor.
The overall evaluation and the statistical measures on all the graded seedlings of the block were also given.