| Autor: |
Bosakova-Ardenska, Atanaska, Kutryanska, Magdalina, Boyanova, Petya, Panayotov, Peter |
| Předmět: |
|
| Zdroj: |
AIP Conference Proceedings; 2022, Vol. 2570 Issue 1, p1-9, 9p |
| Abstrakt: |
This paper presents an application of images segmentation with Statistical Region Merging (SRM) algorithm and median filter for Bulgarian white cheese in brine structure evaluation. Coefficient of segmentation is calculated as a ratio between number of colors in original image and number of colors in segmented image, in order to evaluate how complex is processed image. The effect of kernel size by median filtering as a preprocessing step before segmentation with SRM algorithm is explored using statistical analysis of values of coefficient of segmentation calculated for images of cheese samples which are filtered with median with kernel sizes from 3x3 to 17x17. All examined cheese samples are evaluated by experts according their quality of cut surface and structure. Based on analysis of segmentation results a kernel size 15x15 is preferred for preprocessing step before images segmentation with SRM algorithm. It is performed a correlation analysis for segmentation coefficients versus experts' estimation and the result confirms that 15x15 is preferred kernel size (the correlation coefficient is the highest for this kernel size – about 0,86). This significant correlation is premise for future work over development of intelligent software assistant for white brined cheese quality evaluation. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
