Surveillance and management of parking spaces using computer vision

In this document an algorithm is proposed to identify the state (available/occupied) of the parking spaces in outdoor areas. The algorithm was developed based on two features: the average local entropy, and the standard deviation of the average entropies of subregions of each parking space. The algorithm delivers a binary map, which contains the number of each parking space with its attributes such as area, position and label. The dispersion of the histogram (entropy) is an important factor to extract the information from the frames, since this allows to know if there is uniformity in gray values when there are or there are not any parked vehicles in the parking spaces. With the entropy, it is possible to calculate the two main features posited in this project. A Support Vector Machine (SVM) is proposed by using a linear kernel in order to ensure detection of vehicles.