Design of a Configurable ATR System Using MATLAB

Abstract

Highly efficient and adaptable image recognition systems have become a high priority within the last decade due to terrorism. This paper describes a design and development process for a configurable automatic target recognition system. Our novel template matching system employs a Matlab algorithm developed to accurately detect object patterns within a JPEG image. After downloading the original JPEG image, the Matlab algorithm processes it in three sequential phases: 1) "Sobel" edge detection of the original image 2) Grey- level template matching based on the squared Euclidean distance theory and 3) Statistical pattern recognition of the resulting convoluted image. For our test cases, the cross correlation was determined using a template library created for the algorithm's image recognition process. The average runtime for our system ranges from 5-8 minutes, per test case, with 100% correct recognition using the template matching technique described.