Master of Science
Finding parking in a busy city has been a major daily problem in today’s busy life. Researchers have proposed various parking spot detection systems to overcome the problem of spending a long time searching for a parking spot. These works include a wide variety of sensors to detect the presence of a vehicle in a parking spot. These approaches are expensive to implement and ineffective in extreme weather conditions in an outdoor parking environment. As a result, a cost-effective, dependable, and time-saving parking solution is much more desirable. In this thesis, we proposed and developed an image processing-based real-time parking-spot detection system using deep-learning algorithms. In this regard, we annotated the images using the Visual Geometry Group (VGG) annotator and preprocessed the dataset using the image contrast enhancement technique that attempts to solve the illumination changes in pictures captured in an open space, followed by training the model using the Mask-R-CNN (Region-Based Convolutional Neural Network) and Faster-RCNN algorithms. ROIs (Regions of interest) are used later to determine the vacancy status of each parking spot. Our experimental results demonstrate the effectiveness of our developed parking systems as we achieved a mean Average Precision (mAP) of 0.999 for the PKLot dataset and a mAP of 0.9758 for custom datasets. Furthermore, as part of the smart parking application, we developed an Android App that can be used by the end users. Our proposed intelligent parking system is scalable, cost-effective, and to the best of our knowledge, it offers higher parking spot detection accuracy than any other solutions in this domain.
Summary for Lay Audience
The recent advancement and growth in the automotive industry have significantly increased the number of new vehicles on the road every year. However, this elevation is creating traffic congestion, which increases pressure on existing parking lots capacity in urban areas. Searching for empty parking spaces is time-consuming and has become a major problem in a busy city. Many research works have been developing an intelligent parking spot detection system to address this issue. Most of these research works considered sensor-based solutions. These systems provide accurate results but are expensive to implement and need ongoing maintenance. Besides, the sensors are not very effective in extreme weather conditions in an outdoor parking environment. On the other hand, image-based machine learning and deep learning models have shown the potential to be cost-effective techniques.
This thesis introduces an image-based smart parking solution with deep learning techniques for a real-time parking detection system. This system can detect available parking spots more efficiently and accurately than existing solutions. Our proposed solution and developed prototype can be implemented in real-life parking spot detection, offering cost savings and convenience for the parking lot operators and end users.
Sharma, Aakriti, "Towards the development of a cost-effective Image-Sensing-Smart-Parking Systems (ISenSmaP)" (2022). Electronic Thesis and Dissertation Repository. 8873.