Comparative Analysis of Machine Learning Algorithms in Vehicle Image Classification

Object identification in the area of Computer Vision essentially focus on how robots are able to perceive in comparison of human eyes. The main task for vehicle detection is detecting vehicles from image sources. This study utilises Weka to categorise vehicle photos into 3 vehicle groups based on data mining approach. In feature extraction, three image filtering algorithms are used, which are Colour Layout, Edge Histogram, and Pyramid Histogram of Oriented Gradients (PHOG). The features extracted are then fed into five classification algorithms for a comparative analysis: Multilayer Perceptron (MLP), Simple Logistic, Sequential Minimal Optimization (SMO), Logistic Model Tree (LMT), and Random Forest. The experimental results showed that PHOG filtering algorithm produced the best set of features to accurately classify the vehicle images. During classification, the SMO classifier achieved the best accuracy of 46.3158%.