Quality Evaluation of Edible Beans
A sensing system was developed to evaluate the quality of beans, some of the main characters of the system was to be able to detect
- Color difference
- Cracks
- Foreight objects and broken beans
Crack detection techniques
Results
-
Overall accuracy of 94%(187/200) for fluorescent techniques as compared to 71% obtained for dyed beans using conventional light imaging
-
Laser system provided 98% accuracy in detecting cracks
Self-organizing map combined with a fuzzy clustering for color image segmentation of edible beans
objectives
To efficiently detect beans (acceptable, broken, small and damaged), highly undesirable foreign materials, and stones in a white background
Results
A Kohonen layer of size 32x32 with a "rectangular" topology and a "bubble" neighborhood function, an average of 99.31% of the pixels of the six groups were correctly binarized. The spatial thresholding segmentation procedure provided an average PCMP (percentage of correctly matched pixels) of only 89.71%.
Integrated Product
