The experiment was addressed issue of detecting and classifying landmines at different burial depths. Most landmine detection techniques failed in detection of certain landmine types such as metal or non-metal. Furthermore, several detection techniques necessitated presence of operators in minefields which lead to major threat to their lives. Therefore, there was high demand for reliable systems that required no operator to enter the minefield. In this experiment, thermal image technique was proposed based on landmines had different temperature signature compared to bare soil. The thermal image time series was suggested for better observations. The experiment aims was to design and model a landmines detection system used thermal imaging. The spatial filter was implemented and used with principle component analysis techniques. The spatially filtered image was classified used quadratic discriminant analysis classifier. The classification results indicated performance differences between two classifiers. K-nearest neighbor (KNN) classifier showed superior results compared to quadratic discriminant analysis (QDA). After segmentation, final system output showed nine landmines were successfully detected. The missed landmine was glass cased landmine which buried at 9cm. The burial depth might possible interpretation for missed detection. In additions, other landmines with higher burial depths was detected such as wooden case landmine was buried at 10.5cm.
Volume 12 | Issue 2
Pages: 518-525
DOI: 10.5373/JARDCS/V12I2/S20201073