Document Type
Honors Thesis
Abstract
Pollution is a serious issue faced by the modern world. An analysis of the last five decades reveals that even though considerable attention has been focused on managing larger institutional sources of pollution, individual and crowd-oriented sources of pollution in the form of littering continue to be a pervasive problem worldwide. The 2020 Institute of Industrial and Electronics Engineers (IEEE) region 5 robotics competition was created with the purpose of utilizing robotics to help combat this particular issue of littering. With the goal to represent The University of Texas at Arlington (UTA) at the competition, our senior design team developed an autonomous robot that can recognize various litter items and safely collect and dispose them of in the appropriate bins. This work represents the research and implementation methodology used to integrate computer vision capabilities in the robot. It follows the comparative study of various object recognition models and optimization techniques, a thorough analysis of the team’s hardware and resource constraints, and process of data collection and preprocessing. In addition to mere object recognition, object classification was carried out on a continuous, real-time video stream generated using a high definition (HD) Logitech C270 camera mounted on the robot. This video stream was decomposed to foundational frames and the chosen frames were processed by the object detection model to recognize pieces of litter. The identified litter object would subsequently be labelled as belonging to one of the four approved categories: bottles, chip bags, soda cans and paper trays. A bounding box is created in the video frame to identify the location of the litter and this information can be further utilized to navigate the robot to collect the item. Accurate classification also allows the robot to systematically store different categories of items by potentially integrating compartmentalizing hardware. This systematic storage opens pathways to elevate the autonomous robot’s functionality from basic collection/disposal to proper recycling of litter.
Publication Date
5-1-2020
Language
English
License
This work is licensed under a Creative Commons Attribution-NonCommercial-Share Alike 4.0 International License.
Recommended Citation
Tiwari, Rajvi, "REAL-TIME OBJECT DETECTION FOR RESOURCE-CONSTRAINED AUTONOMOUS ROBOTS TO CATEGORIZE WASTE MATERIALS" (2020). 2020 Spring Honors Capstone Projects. 35.
https://mavmatrix.uta.edu/honors_spring2020/35