Fuzzy Segmentation and Feature Extraction for an Efficient Identification of Mine-like Objects

An efficient landmine detection system is one that should offer a high probability of detection and a low false alarm rate irrespective of the size, shape, explosive, casing, soil in which it is buried, depth of burial and varying environmental conditions. Though 100% probability of detection is still a research question, several techniques are being introduced for an effective demining exploiting various available sensors that are used in landmine detection. In this paper, we focus our attention on finding and evaluating appropriate landmine specific features from the fuzzy segmented infrared (IR) landmine images in order to differentiate them from the background clutters. The system consists of four stages. In the first stage, the acquired IR image is preprocessed using a Gaussian filter to remove the noise and smooth the image. In the second stage, fuzzy segmentation is done on the preprocessed image and the false segmented pixels are removed by post processing the segmented image using various morphological operations. Various structural and Gray Level Co-occurrence Matrix (GLCM) statistical features are extracted in the third stage. A fuzzy inference system is presented in the fourth stage which evaluates the extracted data features and generates a mine confidence value which can be compared to the user defined threshold in order to classify the potential targets as mines or clutters.