Soil contamination classification based on ground penetrating radar data using support vector machine / Mohamad Adib Abdullah

Many of natural sources had been polluted such as water, air, sound and soil(mineral). This research is for making the classification of soil contamination with uncontaminated soil for sand and laterite type of soil. The contamination will be use is formed by hydrocarbon compound which was diesel. This research will be conducted at test bed of GPR scanning in UiTM Perlis. After the data of contaminated and uncontaminated soil are collected, the raw data need to process using Reflexw. The preprocessing of data radar gram consists of move start time, dynamic correction, and hyperbola fitting. GPR data interpretation can be use for classify the buried feature by using machine learning. In this research the classification method that will be using Support Vector Machine (SVM) classifier. The open source provided SVM function is Waikato Environment for Knowledge Analysis (Weka). The SVM classification provided a good quality of classification. All of three soil type classification produce correct instances classified above than 95%. This classification also had been compared with logistic regression classification. The root mean square of these classification provided good result all of them were below 0.05.