Enhanced Deep Convoluted Model for Parking Slot Identification in Smart Cities

Abstract

Finding a parking lot in the metropolitan city is very difficult nowadays due to traffic congestion produced by transportation infrastructure and car parking facilities developed, we're unable to cope with the vehicle congestion on the road. In this proposed work, the parking lot is continuously monitored by using an intelligent CCTV camera. The video is split into the frames of an image and segmented into small parts with the help of the deep convolution neural network. By using a separate framework for deep learning, it is made possible to build a very low-cost system compared to the other state-of-art techniques and also to provide better accuracy and loss function for different climatic conditions.  The main problem in the existing system is accuracy and high cost. Accuracy depends on the climatic conditions, such as haze, exposure to light, poor lighting on rainy days. In the earlier works, most of the developed algorithms can work well for particular climatic conditions only. This solution is cross-validated with PKLot and CNRPark-EXT visual datasets. The overall implementation was performed using a Deep learning studio, and the accuracy produced is 99.5%. The performance of the proposed CNN architecture on the parking slot identification is compared with a different state of the art networks.