A simple digital imaging method for dirt detection on eggshells

The objective of this research was to develop an off-line vision system to detect defective eggshells, i.e., with dirty eggshell, by employing a classification algorithm based on a few logical operations, allowing a further implementation in an on-line grading process. In particular, this work was focused to study the feasibility of identifying and differentiating dirt stains on brown eggshells caused by organic residuals, from natural stains, caused by deposits of pigments. Digital images were acquired from 384 clean and dirty brown eggshells by employing a CCD camera endowed with 15 monochromatic filters (440-940 nm). Each dirty eggshell presented only one kind of defect, i.e., blood stains, feathers and white, clear or dark faces, while clean eggshells did not present organic residuals or evidences of feather, but their external color was characterized by clear or dark natural stains. A MatLab® devoted code was developed in order to classify samples as clean or dirty. The program was constituted by three major steps: first, the research of an opportune combination of monochromatic images in order to isolate the eggshell from the background; second, the detection of the dirt stains; third, the classification of the images samples into the dirty or clean group. The proposed classification algorithm was able to correctly classify near 93% of the samples. The robustness of the proposed classification was observed applying an external validation to a second set of samples, obtaining similar percentage of correctly classified samples (92%).