Identification of impurities in harvested grain using image processing technique

Edwin, Ernieca (2020) Identification of impurities in harvested grain using image processing technique. [Project Paper] (Submitted)

Text lp FK 2020 6.pdf Download (2MB)

Abstract

Paddy (Oryza Sativa) is one of the most important crops in Malaysia. In Malaysia, when it comes to grading paddy, paddy factories such as BERNAS usually receive paddy from farmers. But the paddy they received is often mixed with foreign matters, making it harder for grading process. Using manpower to detect foreign matters in paddy is time-consuming and needs an experienced grader. In this era of Artificial Intelligent 4.0, technology has been used widely including in the paddy industry. In this project, image processing and neural network were used to detect foreign matters in harvested grain. Colour (RGB and HSV) and morphological features (area, perimeter, major axis, and minor axis) were chosen as the parameters to differentiate between paddy and foreign matters. Including paddy, there were 9 different classes predicted, which are paddy, dead paddy, immature paddy, rice kernel, husk, branch with paddy, branch, grass, and others. The result showed that combining both colour and morphological features gives higher accuracy. Quadratic SVM model was chosen as the model to go through the validation process. The validation process has successfully produced an accuracy of 90.66%. However, classes like husk still give a high percentage of misclassification. In the validation process, the misclassification of husk is 38.64%.

Item Type:	Project Paper
Faculty:	Faculty of Engineering
Depositing User:	Ms Siti Mariam Giman
Date Deposited:	21 Nov 2022 07:47
Last Modified:	21 Nov 2022 07:47
URI:	http://psaspb.upm.edu.my/id/eprint/596

Actions (login required)

View Item