Buried Magnetic Material Detection System: An SVM Algorithm Application

Abstract

In recent times, studies about remote-sensing methods have focused on improving variables like sensing distance, sensitivity, and power consumption of available remote-sensing methods. The main target of this study is to improve these variables for the magnetic anomaly detection (MAD) method, in addition to determining upper surface geometry of the sheath with the support vector machine (SVM) algorithm to allow detection of buried ferromagnetic-sheathed explosives, minimize the number of false alarms, and thus shorten the scanning time. The classification process with the SVM algorithm separates this study from others. In line with this, instead of an external magnetic field source, in this study, the Earth's magnetic field was used with a TE100 fluxgate sensor, which has very low power consumption, used as a magnetic sensor to develop a highly sensitive new magnetic measurement and detection system. The system identifies anomalies in the vertical component of the Earth's magnetic field due to the magnetic characteristics of the buried sample in order to determine both the coordinates of the buried object within the scanning field and the upper surface geometry. The system is formed using a five-axis movement CNC platform including sensors and camera. Software was developed using python programming language with a QT graphical user interface (GUI) in order to control the CNC platform, take photographs of the soil, and process sensor data to ensure identification of the upper surface shape of the object under the soil. After obtaining all data for a 30×30 cm <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> scanning area, the software colors the data from blue to red and visualizes this transparently above a photograph of the soil on the monitor. After this stage, a model is created using thresholding, selective search algorithm, and SVM in order to identify the location and upper surface shape of the sample located within the measurement region. Tests performed after SVM training determined the system can classify the upper surface geometry of buried samples with 88% success.