Framework for Outdoor Parking Detection Using Hybrid Model and Its Implementation

In this era of modernization, there is urgent need of automation in the field of detecting vacant lots in the outdoor parking area. As the number of vehicles increasing, particularly in urban areas, people find it difficult to search a vacant lot to park their vehicles. They keep on revolving around the parking area until they get the empty slot for parking or they have to exit if no empty slot is available in the parking lot. In this process of finding a vacant slot in the parking area, people face lots of problems like wastage of fuel, time, mental stress and frustration. In order to overcome these problems, we can think of a system at the entry point of the parking area which shows the available vacant slots in the parking area based on real-time systems. It means, people need not to find empty slots manually but it will be displayed at the entry point of parking area. Therefore, an optimized Parking Management System (PMS) is required which is based on realtime image processing techniques. Hence, in this research, a framework for outdoor parking detection using hybrid model is designed. The primary objective of this research is to evaluate more advanced pre-processors, feature extractors, machine learning/deep learning algorithms for classification of vacant or occupied slots. In this study, the standard, publically accessible dataset ＂PKLot＂ is utilized which consists of 12,417 images of parking lots and 695,899 manually segmented images of parking spaces. The images are captured from parking areas of 4th and 5th floor of the Federal University of Parana (UFPR), and from the 10th floor of the administration building of the Pontifical Catholic University of Parana (PUCPR), Curitiba, Brazil. Therefore, three different sub-datasets namely, PUCPR, UFPR04 and UFPR05, are available in the dataset. The images are captured under different weather conditions like sunny, rainy and cloudy. Hence, in every subdataset folder, there are sub folders namely, sunny, rainy and cloudy. Various challenges in the dataset are studied like noise due to shadows of trees or lamp posts, dim light in rainy or cloudy weather, occlusion of other objects, etc. In order to overcome such noises, pre-processing step plays a vital role. Various image enhancement and image segmentation techniques, available in the literature, are presented. Based on the literature review, we applied CLAHE algorithm for image enhancement and Saliency method for image segmentation to extract ROI.