Help me park !

Abstract

In urban areas, finding a vacant parking slot in parking garages or parking lots is time-consuming and a tedious task for drivers. A system to detect available parking spaces to route drivers efficiently to proper lots is desirable. Some systems have reached the market or are under research promising to support the driver by locating a vacant parking lot. Help Me Park - a video-based system offers a proper alternative to deal with the classification problem. It is possible to combine low-cost hardware requirements with providing detailed occupancy maps for parking areas, which most of the current systems do not provide. Several image processing and machine learning algorithms including Artificial Neural Network, Decision Tree Algorithm, k – Nearest Neighbor, are employed to classify parking slots. By using video-based systems several challenges occur especially on outdoor car-parks. Different weather and lighting conditions or objects occluding parking lots might influence the accuracy for the given task. Problem of shadow is tackled by using edge-detection technique. Help me park refreshes the state of parking lot every 7 – 9 seconds so users are always served with up-to-date information. Depending on the available training data, Help Me Park is capable of classifying slots with 94% accuracy.