A sensor fusion model for the detection and classification of anti-personnel mines. (27th August 2009)

• Record Type: Journal Article
• Authors: Nath, Baikunth; Bhuiyan, Alauddin
• Is Part Of: International journal of innovative computing and applications. Volume 2:Number 1(2009)
• Page Start: 45
• Page End: 59
• Publication Date: 2009-08-27
• Subjects: anti-personnel mines -- APM detection -- ground penetrating radar -- GPR -- infrared imaging -- sensor fusion -- neural networks -- buried mines -- data acquisition -- signal processing -- image processing -- detection accuracy; Evolutionary computation -- Periodicals; Neural networks (Computer science) -- Periodicals; Genetic programming (Computer science) -- Periodicals; Biologically-inspired computing -- Periodicals; Swarm intelligence -- Periodicals; Quantum computers -- Periodicals; 006.3
• Languages: English
• ISSNs: 1751-648X

Abstract

The ground penetrating radar (GPR) and infrared (IR) imaging have become two established sensors for detecting buried anti-personnel mines (APM) which contains no or a little metal. This paper introduces the GPR and IR techniques briefly and describes particular situations where each technique is feasible. We discuss the GPR and IR data acquisition, signal processing and image processing methods. This paper discusses the strengths and weaknesses of each of the sensors based on data capturing efficiency, overcoming environmental difficulties and sensor technology. By providing comparison of these two sensors, we emphasise the necessity of fusion to harness the advantages of each of the methods. We propose a geometrical feature-based sensor fusion framework, combining GPR and IR, as an effective technique for detection and classification of APM which reduces the false alarm rate significantly. We consider the basic geometrical shape descriptor features of an object and construct a feature vector for each of the objects. These feature vectors are used to train a probabilistic neural network (PNN) for the classification of APMs. The method gives almost perfect detection accuracy.