Development of Automatic Target Recognition and Tracking Algorithm with Novel Detection and Classification Approaches
Abstract
Providing an accurate combat identification capability requires an integrated architecture that includes sensor systems, image processing, automatic target recognition (ATR) and target tracking. The goal of ATR is to support rapid and reliable detection, geolocation, tracking, recognition, and prioritization of targets. Image preprocessing, preliminary detection of regions/targets of interest (ROI), classification, and tracking are the four primary components of an ATR system. Applying and collecting data/images in real time using multiple types of sensors simultaneously, identifying the targets from across all sensor output, and fusion of the outputs are critical in making accurate decision. In this project, an ATR system will be formulated with new detection and classification schemes as two major/basic scientific developments. The first basic development will be formulation of a new correlation filter, by combining the concepts of matched filter correlators (MFC) such as extended maximum average correlation height (EMACH) filter, and joint transform correlators (JTC) such as fringe-adjusted JTC (FJTC) for the first time. In EMACH filter, the filter is generated in the Fourier domain from the training images, which is relatively complicated, and correlation is performed in the Fourier domain. In FJTC, the filter or reference image is generated from the training images in the spatial domain following an experimental/ad-hoc procedure, and the test image and reference image are jointly Fourier transformed to determine the correlation output. The proposed correlation filter will be generated in the spatial domain supported by solid mathematical foundation as in EMACH, while the correlation will be performed in the Fourier domain similar to FJTC. This new correlation filter, will work for real time detection of potential targets (i.e. ROIs). The second basic development will be formulation of a novel classifier in light of support vector machine (SVM) and minimax distance transform correlation filter (MDTCF). SVM generally classifies data/samples by constructing hyperplanes and allows classification of data into multitudes of classes. MDTCF attempts to make the filtered false class compact and achieves better separation between the two desired classes. The proposed new classifier is expected to provide real time shift invariant classification of a large variety of potential targets with more accuracy. While developing the novel ATR system with the above mentioned components, some additional achievements are expected; such as determination of appropriate preprocessing mechanism, and identifying proper tracking methodology from the existing methods with needed improvements. The development of the proposed filters will require several important mathematical formulations/manipulations in the areas of vector/matrix algebra, calculus, Fourier optics, and computational intelligence. On the other hand, the work will involve various signal and image processing algorithms, optical correlation, and decision fusion. Besides mathematical and scientific analyses, a significant amount of simulation will be required in the project with Matlab or similar programming languages. All these together will enhance knowledgebase of the involved graduate and undergraduate students at Tuskegee University. A new laboratory on Image Processing can be initiated from the activities of this research, where students will be able to work in the fields of image/video processing, pattern recognition and target tracking. The activities of this research and involvement of the students will facilitate offering a course on Digital Image Processing, which is very attractive in the current era of multimedia. The research and related educational activities are expected to boost enrollment and retention in the department.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Feb 14, 2019
- Source ID
- W911NF1810479
Entities
People
- Sharif Bhuiyan
Organizations
- Army Contracting Command
- Office of the Secretary of Defense
- Tuskegee University