부산대학교 도서관

A camera is a tool to record visual footage in the form of photographs, film or in video format. However, a smart camera can be recognized as a device to retrieve application-specific information from the recorded footage. In this paper, we have proposed a solution to detect parking lot occupancy status using deep learning model and commercially used CCTV cameras in real time. Our implemented solution is decentralized and efficient in terms of light-weight deployment to low powered devices like Raspberry Pi. Our proposed solution is compared with the existing approaches. Our deep learning model is also tested on other datasets having images taking from multiple CCTV camera implemented in different height. Along with this, we have tested our model on indoor both outdoor parking garages in low light conditions during day and evening. Result of the performed experiments shows that our model is operable in low-powered embedded devices with effective accuracy.