SUMMARY

To gain a better understanding of the impact of parking spaces on traffic congestion, road infrastructure, time spent searching for parking and ultimately CO2 emissions, several parking surveys have been undertaken. Traditionally these surveys were performed by individuals manually logging parking lot information, thus limiting spatial and temporal coverage due to high labor costs. As a result, the need for automating the data collection process for open parking spaces is paramount when one starts looking into using parking lot information in real-time applications and across a large area such as a city, state, or even an entire country. Some studies have been conducted looking into this automation process, such as using ultrasonic sensors, 2D LIDARs, vision sensors, parking lot cameras and sensors for individual parking spots in a lot. Most of these methods generated promising results, but were either expensive or not suitable for real-time processing.

This thesis describes an affordable method to detect parking spots in real-time by installing a low cost, off-the-shelf laser range sensor onto a probe vehicle. Several algorithms will investigate the effect of having a stationary sensor and enabling a vertical motion of the sensor, with hopes of obtaining 3D images of a parking lot. The experiments are conducted at different speeds and for different parking configurations, in normal and adverse weather conditions. The results were compared to onboard ground truth camera recordings of the experiments. Statistical analyses were also performed to determine how effective a laser range sensor is in mapping a parking lot in one run, and how many runs were necessary to map the parking lot with a certain confidence.