Automatic separation of defective eggs from quality eggs is an important issue in that both economic and health reasons in the egg farm industry. In the modern egg processing plants, removing defective eggs with visual inspection of people slow process.
Nowadays, different defect detection algorithms are done as well as eggs classification is done with people visual inspection in the real farm industry. These studies are mostly performed on gray level egg images using different types of image processing techniques and decided whether quality egg or defective egg.
In this thesis study, the data set has been produced with photographed producing eggs in natural farm environment. Image processing techniques have been applied on dirty, cracked, broken, bloody and clean eggs within the data set. Firstly, K-means classification technique has been used in the step of separation of color egg image from ground. And then, dirt on the dirty egg has been separated with the color segmentation technique and decided with using threshold value that this egg is dirty. If there isnt any dirt on the image, separated egg from ground has entered second control and Sobel edge detection algorithm by applying egg has been found in parts of the edges and cracks. Edges are dispelled by mask. Then again, it has been decided whether crack egg or clean egg by given the threshold value.
Key Words: Defect detection, K-means classification; color recognition; eggshell defect; visual inspection; image processing.