Detecting landmine fields from low-resolution aerial infrared images

This chapter primarily addresses the problems of landmine field detection through the use of low-resolution infrared (IR) images captured from airborne or vehicle-borne passive IR cameras. We describe a scale-space-based scheme for detecting landmine candidates, as indications of landmine fields. The scheme contains two parts: in the first part, a multi-scale detector, using a special type of isotropic bandpass filter, is employed. In the second part, refinement of landmine candidates is performed through a post-processing scheme that seeks maximum consensus of corresponding landmine candidates over image frames. The chapter also briefly addresses the problems of landmine detection from high-resolution IR images measured at close distances to ground surfaces. A detector based on the thermal contrast model is described. Experiments were conducted on several IR image sequences measured from airborne and vehicle-borne cameras, and on IR images measured at close distances to ground surfaces, where some results are included. Our experiments on these methods have shown that landmine signatures have been significantly enhanced after the processing, and automatic detection results are reasonably good. These methods may therefore be potentially employed for assisting humanitarian de-mining work