Enhancing the development of Android vehicle license plate recognition system via mobilenet SSD.

Malaysia has high density of shopping centers in the city. One famous spot for shopping is located at Petaling Jaya. It has few famous malls within a 10-km radius in the neighbourhood. Generally, about 10 to 30 percent of the visitors drive their own vehicles to the malls. This culture leads mall operators to the situation where car park become an indispensable facility for their malls and the traffic management inside car park an important issue to be discussed with. The bad attitude of Malaysian drivers can be seen by the increase of 22% of compound issued by city council in Kuala Lumpur. The inability to find and contact vehicle's owner in case of emergency incident would negatively impact the safety of visitors and cause traffic congestion in a car park. In order to solve this issue and assist car park management, a system to provide parked vehicles' information to authorized personnel is proposed. In this proposed system, a license plate recognition application has been developed for users to retrieve vehicles' information or register new vehicles by scanning at the vehicles' plate. The application uses MobileNet Single Shot Detector (SSD) algorithm to read license plate and query for vehicles' information from Firebase using the unique combination of plate number. In this paper, MobileNet SDD is compared with other light object detection algorithms such as You Only Look Once (YOLO) and other variants of SSD for their strengths and weaknesses. The vehicles identification methods used in car park are compared to show the trend of increasing use of License Plate Recognition (LPR) systems in shopping centres. The project is an attempt to take advantage of the LPR technology trend through proposing an extension to such a system on mobile device. This helps car park operators to get the owners' information without the hassle to manually key in details at a fixed location system. [ABSTRACT FROM AUTHOR]