Data Fusion For The Detection Of Buried Land Mines

Author : Mohammed Hussein Baydoun
Publisher :
ISBN 13 :
Total Pages : 462 pages
Book Rating : 4.:/5 (787 download)

DOWNLOAD NOW!

Download or read book Protocol for Detection and Classification of Buried Landmines Using Infrared Imaging written by Mohammed Hussein Baydoun and published by . This book was released on 2007 with total page 462 pages. Available in PDF, EPUB and Kindle. Book excerpt: Landmines are a worldwide problem and their detection is an active field of rese arch whether the mines were metallic or not. Infrared imaging is considered as t he best single sensor used to detect shallow buried landmines. The strength of I R imaging is due to the existence of the thermal signature of the buried object at the soil surface due to the different temperature variations induced by burie d objects. We aim at building a protocol for detecting and classifying buried landmines. Fo r this purpose, we apply image processing techniques on IR images due to the the rmal signature property of buried objects in general and landmines in particular . The work focuses on one type of landmines but can be easily extended. In the demining field, it is established that one indicator is not enough but ra ther a system of multiple sensors fused together is a must to achieve a capable system. Thus, the IR imaging system is combined with a direct normal vision algo rithm to remove surface laid objects. Also, it is enhanced using different scena rios that include water spraying with the day and night scenarios. A special setup which consists of a soil bed with different sensors and a weathe r station is prepared to obtain necessary data whether IR images or any variable like specific temperatures or radiations. This is an outdoor but not a field se tup and it is based on an indoor soil bed previously prepared according to (Kabb ani 2005) and (Khalid 2005). The different experimental data is compared with a three dimensional numerical model developed by (Saleh 2004). The three dimension al model is reduced into a one dimensional one for speeding the process and that model is enhanced to match the three dimensional model with a reduction of erro r up to 60%. Furthermore, two databases of synthetic thermal signatures based on the numerica l simulation data are constructed to achieve mine classification. The first data base is small and restricted to the soil type used in experimentation with 240 r ecords while the second one is large and extends to 473 soil types in the vicini ty of the soil type used. Only the small database space is searched using differ ent techniques that include genetic algorithms and neural networks.