Pine nuts selection using X-ray images and logistic regression
Ikramullah Khosa, Eros Gian Alessandro Pasero · 2014
Automatic and quick evaluation of ingredients as well as end products is getting more attention in recent days in the food industry, so as to make the production process fast and efficient. In this paper, binary classification of pine nuts using x-ray images is presented. Independent nutmeat x-ray images are extracted and two kinds of features are extracted from them. A set of features is produced projecting statistical texture properties of the images, using Gray Level Co-occurrence Matrices (GLCMs). Edge detection is applied, histograms of images after edge detection are produced and used as second and third set of features with 40 and 30 bins respectively. Eighty percent of examples from each; good and bad category, are used for training purposes and rest are used as test data. Logistic Regression with gradient descent algorithm is used as classifier. Results are calculated as classification accuracy, sensitivity and specificity. The classifier produced better results with simple features achieving maximum specificity in comparison with similar solutions for such classification problems.