Comparison between k-Nearest neighbors, self-organizing maps and optimum-path forest in the recognition of packages using image analysis by Zernike moments

Recognition of objects using an industrial image sensor is an important tool that has been motivated by the necessity of automatic recognition systems in the industrial automation. In this context, an interesting problem is the automatic image acquiring and a high reliability in objects classification. To this end, this paper presents a comparison between k-Nearest Neighbors Classifier using Euclidean, City Block, Cosine and Correlation distance metric, the Self-Organizing Map (SOM) - Artificial Neural Network (ANN) and the Optimum-Path Forest, for classification of images taken from a low-resolution industrial sensor. Classification performance has been compared in terms of extraction time and accuracy using image analysis by Zernike moments.