A NEWTONIAN REPULSIVE FORCE FIELD AND MORPHOLOGICAL SKELETON PARAMETRIC CURVATURE NEURAL NETWORK RECOGNITION SYSTEM FOR HUMAN DETECTION AND TRACKING WITHIN A LAGRANGIAN SYSTEM
Open Access
Author:
Richards, Ryan John
Area of Honors:
Electrical Engineering
Degree:
Bachelor of Science
Document Type:
Thesis
Thesis Supervisors:
Dr. Timothy Joseph Kane, Thesis Supervisor Dr. Julio V. Urbina, Thesis Honors Advisor
Keywords:
human detection tracking infrared camera morphological skeleton newtonian potential field force field divergence curvature ricci flow raspberry pi OpenCV
Abstract:
Human detection and tracking is a growing concentration within image processing and electrical engineering that possesses many applications, including reconnaissance, weapon interface, search and rescue missions as well as autonomous vehicles. This thesis investigates two feature extraction methods, unprecedented in the detection field: (1) the Newtonian repulsive force field and (2) dilated morphological skeleton’s parametric curvature. The extracted information is classified using an artificial neural network that has been calibrated from situational-based images. Constituents of the proposed system are demonstrated to be efficient and mutable by implementing them in multiple coding languages and ultimately deploying them to a Raspberry Pi for testing. By a robust but simplified algorithm, this thesis has concluded that the proposed object descriptor provides around 80% accuracy (contingent upon response bound definition) and has demonstrated versatile properties, such as accurate responses under non-situational calibrations.
