Adaptive Change Detection Methodology for Buried Mine and IED Detection from Space

*† The paper describes an image processing methodology developed to detect buried improvised explosive devices (IEDs) and landmines. The methodology assumes that sequences of aerial images of the same terrain, for instance a dirt road, can be obtained at periodic intervals of time. Corresponding images of the same terrain from two consecutive sequences can be processed to detect if the terrain has changed or been disturbed and may contain buried landmines. The image processing methodology requires registering corresponding images by identifying matching natural landmarks in both images, then calculating and performing scale and perspective transformations so that both images can be superimposed and compared. An adaptive methodology to compare registered images was developed in order to avoid false alarms due to changes in illumination or other small natural changes. The methodology was tested in a laboratory setting using a medium size Cartesian robot to manipulate a charge-coupled device (CCD) camera. A sand and rock box was placed within the robot workspace volume. Several cylindrical disks of controlled sizes and aspect ratios were used to simulate landmines. Several image sequences were obtained before and after burying the simulated landmines. Extreme care was observed in burying the simulated landmines so that the terrain does not look as if it had been disturbed. The results obtained with this methodology are good. The paper will describe the methodology, testing environment, tests conducted and results obtained.