Determination of broken rice percentage using image processing technique
Physical rice grain composition is one of the components used for rice grading which involve the identification of head rice and broken rice. Rice grading is important to ensure only edible rice reaches the consumer standard. It also protects consumers from price manipulation. In this study, a new a...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2014
|
| Subjects: | |
| Online Access: | http://psasir.upm.edu.my/id/eprint/56249/1/FK%202015%2066RR.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Physical rice grain composition is one of the components used for rice grading which involve the identification of head rice and broken rice. Rice grading is important to ensure only edible rice reaches the consumer standard. It also protects consumers from price manipulation. In this study, a new approach of image processing technique has been developed to detect and identify head and broken rice based on its physical properties i.e. area, perimeter, minor axis length and major axis length. The rice images were first segmented automatically from its background by using three different image segmentation methods, namely Mean Iterative,Median Iterative and Otsu's method. The Otsu's method provides satisfactory results. It only needs four iterations to complete the process of detection and identification, meanwhile Mean Iterative and Median Iterative takes 11 and 12 iterations, respectively. Furthermore, Otsu's method takes the average of four seconds to run the experiment, meanwhile Mean Iterative and Median Iterative takes average of 10 seconds to complete the experiment. Connected component analysis was later being applied to eliminate unwanted noise. Results from the statistical analysis have shown that area and perimeter give significant correlations with all of the other properties. However, area gives more consistent result with the value of correlation greater than 0.6 in all properties. Therefore, an area was later used as input parameter in developing a simple model of head and broken rice identification using a linear regression analysis. The model give promising results when tested with 0%,1%,5%,10%,15% and 20% of broken rice taken from 600 samples of rice images with the average percentage accuracy of 98%. Graphical User Interface (GUI) software was developed via MATLAB R2014a to help user directly access the percentage of broken rice from an image. |
|---|