Smart Campus: Carpark Management System Using Video Analytics

by P Y Fung (FMO), Joseph Ng (ESU)


As part of an initiative to build a smart campus, the Carpark Management System at the Bank of China (Hong Kong) Complex (BOC), Yeung Kin Man Academic Building (YEUNG) and Lau Ming Wai Academic Building (LAU) have been upgraded with the video analytics technology.  The new Carpark Management System uses the license plate recognition (LPR) and Octopus systems to link up with the workstations at the entrance and exit terminals of various carparks.  The data records in any one of the workstations will automatically disseminate to all workstations of the system. 

When a car enters the carpark, the new Carpark Management System will use an infrared camera to video capture the car’s license plate and run the video analytics software on it.  The numbers and letters on the license plate are then converted into data which is then matched against the campus parking permit database.  If the license plate is found in the database, the traffic barrier automatically opens.  In case the LPR system cannot recognize the license plate, staff members can still tap their staff cards against the CityU card reader to get access into the carpark.

The system also possesses the anti-passback feature so that no double entry of any user will be allowed.





The benefits of using the LPR and Octopus systems in the new carpark system are manifold. Firstly, the Facilities Management Office (FMO) no longer needs to issue parking cards to staff members who applied for the annual parking permit.  Data input will automatically be uploaded onto the central database after the applications have been approved.  Thus, staff members do not need to keep multiple cards nor roll down windows for card tapping against the card reader, which is especially useful at the campus ground level carpark of YEUNG during adverse weather conditions. 

Secondly, it improves the traffic flow during peak periods.  As compared to the time required for drivers (i.e. staff and guests) to stop the car at the gate and tap the parking card, it only requires 2 to 3 seconds to process the LPR and lift the gate. 

Thirdly, departments are no longer required to issue parking coupons to their guests if prior application has been made to the FMO. They only need to pass the license plate number to the FMO for inclusion in the guest database. This could save a lot of administrative time and effort. 

Finally, visitors could use Octopus cards to settle their parking fees directly at the exit terminal of the carpark.  The use of Octopus saves a lot of hassle in handling cash and administrative staff no longer be required to bank-in the parking fees on a daily basis. 

Technology Behind
The LPR system is a video processing technology, which uses optical character recognition (OCR) on video to read the car’s license plate in real time.  This technology was first invented in 1976.  It did not become widely used until 2000s, due to the revolution of hardware over the past few decades.

To read and recognize the license plate from a video image, the steps involved are as follows:

  1. Image pre-processing
  2. License plate positioning
  3. License plate segmentation
  4. License plate character recognition
Steps in License Plate Recognition
To begin with, the system uses an infrared camera to take a video of the car.  The infrared camera helps to acquire a clear license plate image at any time of the day or night, and in almost any weather conditions.
Before recognizing the license plate, several image pre-processing methods must be used to improve the accuracy of the license plate images from the infrared camera.  For example, smooth filtering to remove image noises.
With license plate positioning, the edge detection method detects the location of the license plate and passes the license plate image to the next step.
In the case of license plate segmentation and character recognition, the segmentation of the character image is extracted from the license plate image.  Then, the pattern of each character is analyzed and converted into text in real time.
Due to the rapid development of cameras, mobile devices and video analytics technology, it should come as no surprise to have high quality cameras and smartphones or tablets of strong processing power to meet the increasing requirements of the LPR video analytics system.  In the future, some mobile applications can even be developed for license plate verification and real time parking information enquiry, etc.  These applications may also recognize the vehicle’s colour, model or brand.  
We would like to take this opportunity to express our sincere gratitude to the Central IT’s unfailing support in making the project a success. The project lasted for nearly 8 months and involved a lot of IT issues during the system design stage, the subsequent work implementation and troubleshooting stage. The Central IT has provided tremendous help and assistance, and offered valuable professional advice to assist us to overcome all the hurdles which included walking through the whole design logic with the project team and the contractor, network installation, configuration of the smart card reader and parking card, problem solving, etc. Without their kind help, the project would not be able to complete on time within such a tight working schedule. We would also like to express our sincere thanks to the Campus Development Office (CDO) in assisting us on various building services and related issues so that all the site work can be carried out smoothly.  
Wikipedia: Automatic number-plate recognition