Synthetic template: effective tool for target classification and machine vision

Kaveh Heidary

doi:10.14569/ijacsa.2013.041005

Abstract

A process for replacing a voluminous image dictionary, which characterizes a certain target of interest in a constrained zone of effectiveness representing controlled states including scale and view angle, with a synthetic template has been developed. Synthetic template (ST) is a spatial map (grayscale image) obtained by combining the set of zone-specific training images that are ascribed to the target of interest. It has been shown that the solo-template ST correlation filter outperforms filter banks comprised of multiple target-class training images. A geometric interpretation of the basic ST concept is employed in order to further explain and substantiate its properties.

Highlights

Machine vision involves the process of autonomous assessment of imagery data in a wide array of applications ranging from robotic navigation to biometrics and automatic detection and tracking of targets [1-5]
The performance of correlation filters based on bank of templates (BT), prototype template (PT), synthetic template (ST), and fractional bank of templates (FBT) are examined using actual images for training and testing
The images of the chosen target-class object which were not employed in the training process as well as the images of other non-target objects were utilized as the test set of images

Summary

INTRODUCTION

Machine vision involves the process of autonomous assessment of imagery data in a wide array of applications ranging from robotic navigation to biometrics and automatic detection and tracking of targets [1-5]. In order to reduce the arduous computational burden of storing and processing vast image dictionaries, arising from the object image variability effects caused by intrinsic and extrinsic inconsistencies, synthetic discriminant functions (SDF) and distortion tolerant filters have been developed [2946]. In order to establish the presence and location of the target of interest in the sensor image, rather than computing the cross correlations of the input image with respect to all the target dictionary images, it is correlated with the single-template ST. This results in storage and processing savings proportionate to the number of images in the original target dictionary.

PROBLEM FORMULATION

TEST RESULTS

GEOMETRIC INTERPRETATION

Image-Point Analogy

Simulation Results

CONCLUSIONS