Image Processing Techniques for Lateral Migration Radiography Land Mine Images

摘要

Lateral migration radiography (LMR) is a new imaging technique based on ideas of Compton backscatter imaging. In LMR, multiple image sets are generated. Uncollimated detectors provide images that are due primarily to single-collision photons which have been mainly scattered from the soil surface or near-surface. Collimated detectors provide images that are due primarily to multiple collision photons that contain both surface and sub%surface information. Simple noise removal and image enhancement techniques, including weighted filters. Weiner filters and histogram equalization improved the visibility of objects and of object details. Linear combinations of collimated and uncollimated images provided a fast, simple means of surface object removal and of yielding additional details for mine detection. Optimal filters that maximize the signal to noise ratio and include a prewhitening step were particularly useful for identifying deep buried mines. Finally. simple neural networks gave good image enhancement results and are especially efficient or surface-laid object removal.