Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
6539976 | Computers and Electronics in Agriculture | 2017 | 11 Pages |
Abstract
In a trading market, price of oil palm (Elaeis guineensis) is negotiated depending some key parameters of the fresh fruit bunch (FFB). Inspectors have been hired by a buyer to grade FFB to accept or reject. The classification results made by human inspection are skeptical and not very reliable if workload is high. We have developed a system to grade FFB depending on its quality. Several palm features are extracted from RGB, near infrared, and depth images, captured with a Microsoft Kinect camera version 2.0 installed in a light-controlled environment on the conveyor line. Two main algorithms for classification have been developed. The first algorithm is called a volume integration scheme (SVIS), which measures the relative volume of palm bunch. The second developed algorithm classifies palm bunch into three grades (L-Grade, M-Grade and H-Grade) based on oil content from Soxhlet extraction. The system achieves 83% accuracy for grading palm bunch within 6â¯s per one sample, which shows the possibility of using the system in a trading market.
Related Topics
Physical Sciences and Engineering
Computer Science
Computer Science Applications
Authors
Burawich Pamornnak, Somchai Limsiroratana, Thanate Khaorapapong, Mitchai Chongcheawchamnan, Arno Ruckelshausen,