Accurate detection of structural edges of a room in a photograph using openCV canny edge detector in python

• Main Author: Tan, Nicholas Jia Long
• Other Authors: Lee Yong Tsui
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2021
• Subjects: Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision; Engineering::Aeronautical engineering
• Online Access: https://hdl.handle.net/10356/149327
• Institution: Nanyang Technological University

This project aims to study the accurate detection of structural edges of a confined room environment using the OpenCV Canny Edge Detection (CED) algorithm. Edge detection is classically well-understood, and many edge detectors exists today. Amongst them, the CED algorithm, developed by John Canny, is arguably one of the most well-known and well-studied edge detectors for its high accuracy and ability to eliminate noise. However, challenges exist in applied use of edge detection technology for specific purposes. In this project, a simple methodology is proposed to breakdown an image into component regions of interests. CED threshold tuning is then performed for each individual region to achieve accurate structural edge detection. Finally, component regions are then reconstructed to produce an overall structural edgemap of the original image. The proposed methodology is evaluated for its effectiveness and efficiency in producing a complete (all visible structural edges being accurately detected) and clean (zero noise and no undesired, non-structural edges) edge-map. Other features available in OpenCV, such as the histogram plot function, were also explored in their efficacy towards achieving the project objectives. Lastly, this report discusses two main issues hindering the success of the project which are interferences to edge detection due to the presence of reflective surfaces and the presence of shadows.