%0 Journal Article
%T Distance independent weight estimation system based on image processing with artificial neural networks: egg and orange samples
%A Umut Engin AYTEN
%A ¨¹nsal Burak £¿ALVARCI
%J -
%D 2019
%X Measuring the weights of the objects in industrial and academic studies is highly important. Therefore, it is aimed to perform weight estimation as independent of distance and camera angle using artificial neural networks (ANN) based on image processing procedure in this study. Feed-forward multi-layer perceptron (MLP) and radial basis function (RBF) network were used as artificial neural network structure. Orange and egg were selected as sample objects to be estimated the weights. To train and test the system with these samples, 250 eggs samples from 4 different brands and classes (x-large, large, medium, and small) and 150 oranges samples with different sizes were selected. Using these samples, a database containing 750 images for eggs and 450 images for oranges taken under the right angle, positive angle and negative angle was created. The weight estimation system consists of a camera, artificial lighting system, reflector and reference image and also the MATLAB program and toolboxes were used during the weight estimation process. In this study, to establish a system with highest success value was aimed by testing the different feature vectors, images taken from different angles and different ANN parameters. The system created by each change was run five times and arithmetic mean of the results was taken. In addition, the experience, which has the highest success value, was also calculated by K-fold cross validation method to obtain the success value. While the accuracy rate obtained from the results weighted by accurate weighting instrument is 47% according to definition of Turkish food codex egg communique and classification values written on the egg cartoon, this rate was increased to 90.5% for MLP, 86.3 5 for RBF in this study. Moreover, the success rates of the weight estimation for MLP and RBF were calculated as 99.42% for eggs, 98.20% for oranges and 99.43% for eggs, 98.25% for oranges, respectively
%K A£¿£¿rl£¿k tahmini
%K Yapay sinir a£¿£¿
%K Radyal tabanl£¿ fonksiyon
%K Say£¿sal g£¿r¨¹nt¨¹ i£¿leme
%K Yapay sinir a£¿£¿
%K £¿ok katmal£¿ alg£¿lay£¿c£¿
%U http://dergipark.org.tr/hkmojjd/issue/45183/538349