Geometric and Pattern Recognition Combined Algorithms Applied to Digital Image Processing in Aerial Search and Rescue Applications
Date of Award
Thesis - Open Access
Master of Science in Mechanical Engineering
Patrick Currier, Ph.D.
First Committee Member
Eric Coyle, Ph.D.
Second Committee Member
Richard Prazenica, Ph.D.
The difficult task of generating an automated target detection system from aerial imagery can be achieved through different approaches. However, the robustness of any machine vision system lays on the implementation of vision algorithms. By combining geometric and pattern recognition algorithms, a more effective machine vision scheme is developed. Three modules-Color Identification (CIM), Color Matching (CMM), and Pattern Recognition (PRM)- work as main filters to process images and predict the location of targets based on a set templates. Multiple target detection is also accomplished by implementing a hide and seek method where all matching objects are scanned individually and isolated to keep count of the total amount of similar possible targets. Two different scenarios are analyzed in this process: Outback Challenge from 2012 Search and Rescue Competition and Bird counting where Cormorants and Common Murres interact in the same habitat. Results are analyzed in two different groups for each scenario. First group includes individual module analysis, while second group combines all three modules (CIM,CMM, and PRM) in every analysis.
Scholarly Commons Citation
Correa, Jamil Efren Anaguano, "Geometric and Pattern Recognition Combined Algorithms Applied to Digital Image Processing in Aerial Search and Rescue Applications" (2013). Doctoral Dissertations and Master's Theses. 13.