Fruit image classifier based on artificial intelligence

Abstract

Machine vision and image analysis applications are nowadays capable of solving various problems in the industrial,  scientific or safety sectors. Image classification is very useful for the automation of processes in a company. In order to perform an image classification task, the features identifying each kind of image, such as color, shape, and texture, must be extracted. In the present work, it is necessary to implement the algorithms for the construction of a fruit image classifier, based on image color features extraction in certain regions of interest. For developing the fruit image classifier, the threedimensional color histogram extraction technique is used, and with the implementation of artificial intelligence algorithms, image automatic classification is accomplished. The dataset used consists of: four fruit types with a varied number of images per class, then the images are prepared by selecting regions of interest through the use of masking techniques and they are then divided into two datasets: training data and test data. After the classifier is trained, classification tests are performed to evaluate the effectiveness of the fruit image classifier. This classifier implementation and construction methodology can be used in various applications, depending on the types of object images to be analyzed in similar conditions, and to automate classification and object recognition processe