Off-campus UNL users: To download campus access dissertations, please use the following link to log into our proxy server with your NU ID and password. When you are done browsing please remember to return to this page and log out.
Non-UNL users: Please talk to your librarian about requesting this dissertation through interlibrary loan.
A Novel Approach to Low-Resolution Occupancy Sensing Using Dynamic Feedback Comparison on Successive Pixelized Still Images
A NOVEL APPROACH TO LOW-RESOLUTION OCCUPANCY SENSING USING DYNAMIC FEEDBACK COMPARISON ON SUCCESSIVE PIXELIZED STILL IMAGES Arpan Guha, Ph.D. University of Nebraska, 2020 Advisor: Dale K. Tiller, D.Phil Occupancy sensing has been extensively researched over the past three decades and various commercially available sensor technology exist today. Amongst standalone sensors, passive Infra-red (PIR), ultrasonic (UL), and dual-technology sensors are commercially popular because of their cost-effectiveness. However, PIR, UL, and dual-technology sensors have low data resolution and require data acquisition systems to collect and analyze the data obtained. Their inability to detect static occupancy is an often-cited flaw where the sensors time out when there is little or no detected movement caused by occupants in the monitored space. Camera-based systems are possibly the most reliable method of occupancy sensing simply because of the high resolution of data acquired, but they have their share of disadvantages in terms of the cost of the technology and data acquisition system, and most importantly, the privacy invasion aspect. Taking these factors into account, this dissertation describes a method where information extracted from pixelized still images can be analyzed to predict occupancy, and all this has been implemented using an inexpensive RaspberryPi infra-red camera module. The camera module captures images, pixelizes them, converts them into an array of numbers, and analyzes the relationships between successive arrays to characterize occupancy. Pixelizing the images ensures that privacy is preserved while the hardware module collects and analyzes its own data which results in a small form factor and much lesser cost. The proposed method successfully separated signal (human occupancy/motion) from noise (potential extraneous confounding stimuli) and yielded an overall detection accuracy of 97.92% compared to the ground truth, which affirms its potential of being extended to commercial applications.
Guha, Arpan, "A Novel Approach to Low-Resolution Occupancy Sensing Using Dynamic Feedback Comparison on Successive Pixelized Still Images" (2020). ETD collection for University of Nebraska - Lincoln. AAI27959819.