Ground Penetrating Radar Slice Reconstruction for Embedded Object in Media with Target Follow

The detection of embedded object from ground penetrating radar GPR imagery is our goal. The GPR image is a cross sectional slices. The embedded objects are metal and/or plastic type. In many fields demand for visualizing objects scanned as cross sectional slices is growing. This research has many real world applications, such as robotic environments, medicine, remote sensing, inspection of industrial parts and geology. An even better way is to visualize the underground object by reconstruction a threedimensional model of those objects from the slices. Objects here are stable underground while, camera is moving. The task of object track in a cross sectional slices consists of two parts: first gather information on changes between succeeding slices (object detection), and second process this information appropriately to obtain the track of an object. If the object is like cable or pipe. The proposed method starts with two dimensional 2D image preprocessing for each slice. The preprocessing involves multispectral to gray conversion, contrast enhancement, segmenting, thresholding and denoising to modify each 2D image slice individually. Preprocessing algorithms involved in this paper are chosen appropriately to have image without noise, with object detected and with object eliminated. After a preprocessing step the proposed algorithm for object detection starts with objects contour finding in each slice, 2D objects transparency and transformation. The last step is the proposed interpolation technique to build the successive slices until the spaces is filled to find out the embedded object. Key-Words: Object Detection, Object Recognition, Ground Penetrating Radar (GPR), Volume Reconstruction, Interpolation, Image Processing, Target Follow.