Electronic Thesis and Dissertation Repository

Thesis Format



Master of Science


Computer Science

Collaborative Specialization

Artificial Intelligence


Yalda Mohsenzadeh

2nd Supervisor

Michael Bauer


Advances in Computer Vision and Aerial Imaging have enabled countless downstream applications. To this end, aerial imagery could be leveraged to analyze the usage of parking lots. This would enable retail centres to allocate space better and eliminate the parking oversupply problem. With this use case in mind, the proposed research introduces a novel framework for parking lot occupancy assessments. The framework consists of a pipeline of components that map a sequence of image sets spanning a parking lot at different time intervals to a parking lot turnover heatmap that encodes the frequency each parking stall was used. The pipeline of components includes Image Stitching, Vehicle Detection and Heatmap Generation. The focus of this work is Image Stitching and Vehicle Detection, while Heatmap Generation is left for future work. Beyond proposing a novel framework for parking lot occupancy assessments, several contributions are made to the Computer Vision field. In particular, a novel method for initializing the pose of images based on the metadata from the acquisition system is introduced. Additionally, a novel comparative study of object detection models applied to the vehicle detection task is presented. Extensive experiments are used to validate the proposed contributions on both public and private datasets.

Summary for Lay Audience

Vision is fundamental to how humans perceive and act in the world. Thus, in the pursuit of creating intelligent systems, it is intuitive to endow computers with a similar sense of perception. This is the focus of Computer Vision - a scientific field seeking to develop systems that analyze images and videos. In this thesis, Computer Vision is leveraged to work toward a system that performs parking lot occupancy assessments. More specifically, the system uses aerial images of parking lots taken across time intervals to generate a heatmap that encodes the number of times each parking stall is used. The system consists of a pipeline of three components: Image Stitching, Vehicle Detection, and Heatmap Generation. First, Image Stitching is used to map a set of overlapping images to a consistent mosaic. Subsequently, vehicle instances in the mosaic are localized during Vehicle Detection. Lastly, Heatmap Generation uses the mosaics along with the detected vehicle instances to generate a heatmap. The focus of this work is on Image Stitching and Vehicle Detection. However, a preliminary discussion of Heatmap Generation is included to provide some initial direction.