Detection and classification of land military vehicles with machine learning methods

This study focuses on the detection of military vehicles using image processing and machine learning algorithms. Four different categories were selected as military vehicles: BMC 6x6 truck, M113 Armored Personnel Carrier, Land Rover Defender, and Tank. A dataset consisting of a total of 200 images, with 50 images for each type of vehicle, was prepared. In the first step of the study, pre-processing was performed on the images using image processing techniques. The pre-processing phase includes background removal, sharpening filter, and image resizing. Following this step, feature extraction was conducted using Histogram of Oriented Gradients (HOG) and Two-Dimensional Wavelet Transform (2D-WT) methods. The extracted features were subjected to feature selection using ReliefF (RF) and Principal Component Analysis (PCA) to reduce the dimensions of the feature matrix. The obtained features and labels were then applied to Decision Trees (DT), K-Nearest Neighbors (K-NN), Naive Bayes (NB), and Cascade Forward Neural Networks (CFNN) algorithms. Metrics such as confusion matrix, accuracy, F1 score, and precision were used to evaluate the performance of these classifiers. The results showed that the performance of each classifier was separately applied and examined with feature extraction and feature selection methods. It was observed that the CFNN algorithm achieved high accuracy and F1 scores by capturing the complexity and details of the visual data in the best way. The best results were obtained in the HOG feature extraction method and RF feature selection methods. As a result of this study, the effectiveness of image processing and machine learning algorithms in the detection of military vehicles was demonstrated, providing a significant foundation for future research in this field.