Design, Implementation and Comparative Study of Supervised Classification Algorithms for Object Sorting

Abstract

The present study contributes the detailed design, Implementation and comparative studies of supervised classification algorithms such as Nearest Neighbor (NN),k-Nearest Neighbor (k-NN) and Minimum Mean Distance (MMD) along with the distance metrics Maximum, Sum (Manhattan) and Euclidean distances. Classification involves training and testing phases. Training phase teaches the classifier about the types of sample images to classify during the testing phase and saves the class names in a classifier file. Testing phase classifies an unknown sample image into one of the class from classifier file. Total numbers of sample images were divided into the ratio of 30:70 for training and testing respectively. Then classification procedure was performed on sample images for each algorithm with each distance metric. The training and testing VIs for particle and color classification were designed and implemented by using LabVIEW. All the algorithms and distance metrics were analyzed, compared for best results and the maximum accuracy algorithm along with its distance metric is going to implement in real time object sorting application.