Award-Winning Project Sharing: License Plate Recognition SoC Design Based on Arm Cortex-M3 Processor

Follow us by clicking the blue text above

From August 19-21, 2019, the 3rd National College Student Integrated Circuit Innovation and Entrepreneurship Competition (hereinafter referred to as the “Competition”) was successfully held in Nanjing Jiangbei New District Research and Innovation Park. As a cup competition unit of the competition, Arm China set up the Arm SoC Design Challenge.

Since the start of the competition, a total of 197 teams from seven major competition areas across the country registered for the Arm cup competition. After enterprise evaluation and sub-competition preliminary rounds, 27 teams advanced to the national finals. After a day of live demonstrations and evaluations on August 20, the Arm cup competition produced 4 first prizes, 7 second prizes, 10 third prizes, and 5 excellence awards. In addition, the WNS<0 team among the first prize winners stood out and won the Arm cup enterprise special award.

To allow everyone to gain a deeper understanding of the excellent works of the students, promote communication and learning among students, and encourage more students to participate in competitions, Arm China Classroom will share some award-winning team works with everyone.

Beijing Institute of Technology

WNS<0 Team

One

Arm Cup – Arm SoC Design Challenge

Competition Topic: Arm SoC Design Challenge

Competition Requirements: Undergraduate Group/Unlimited Group

Competition Content:

Utilize the Arm Cortex-M3 processor IP provided by DesignStart Eval to construct a system on a programmable logic platform, achieving image acquisition, image processing, and human-computer interaction functions, and observe and optimize the system’s performance.

Two

Team Introduction

Participating Unit: Beijing Institute of Technology

Team Name: WNS<0

Instructor: Gao Wei

Team Members: Luo Haoyang, Liu Zhaojia, Luo Yixuan

Finals Awards: First Prize and Enterprise Special Award

Three

Project Introduction

The license plate recognition system, as an important part of intelligent transportation systems, is now widely used in various aspects of daily life, playing an irreplaceable role in highway tolls, parking lot management, traffic department monitoring, and target detection in autonomous driving. The Arm Cortex-M3 processor has advantages of high performance, low power consumption, and low cost. Therefore, designing a license plate recognition SoC based on the Arm Cortex-M3 processor will have significant advantages in application area and cost-effectiveness.

This license plate recognition system is developed using the Perf-V A7 100T FPGA development board, based on the Arm Cortex-M3 processor, utilizing an OV5640 camera for video acquisition, inputting license plate images into the system, and processing the license plate recognition algorithm with the processor system and hardware accelerator. The output results are displayed on an external LCD screen, and a buzzer indicates that recognition is complete. Meanwhile, the camera’s video will also be displayed in real-time on the screen for comparison. In addition, this design also uses the supporting expansion board provided by the Perf-V A7 100T FPGA development board, integrating the OV5640 camera into the development board, and also uses a self-made Arduino interface expansion PCB board to integrate the LCD screen, buzzer, and rotary encoder into the development board, and designed an automatic page-turning extension device. Through continuous optimization of the system, it can now complete high-speed and accurate recognition of Chinese characters, letters, and numbers in 20 license plates within 1 second.

The hardware part of this design is built using Verilog HDL language on Vivado, and the software part is compiled and debugged using C language on Arm Keil MDK.

Four

System Architecture Diagram

Five

Function Verification

Camera Configuration Function Simulation	Camera Data Input Function Simulation
LCD Function Simulation	Projection Accelerator Actual Test Results
Matching Accelerator Test Results	License Plate Recognition Results

Six

Technical Innovations

1

Using quadtree division for character extraction local features for license plate recognition, compared to traditional projection template matching methods, effectively improves the accuracy and anti-interference ability of template matching methods;

2

Using template offset comparison methods, allowing normal recognition of license plates at small angles of tilt, and using local template matching methods to improve the recognition differentiation of similar characters;

3

By optimizing the algorithm flow, reducing data transfer between DTCM and shared BRAM within the Arm Cortex-M3 processor, while cooperating with three hardware accelerators and a camera refresh rate of up to 120 frames, significantly improving recognition speed;

4

Using a rotary encoder to achieve digital zooming and shifting of the camera output image, allowing for changes in the recognition area without moving the physical location of the system.

Seven

Project Video

Eight

Competition Experience

Competition Feelings

During the competition, we had a very personal experience of the differences in design ideas, development difficulties, and time consumption between software and hardware development, fully experiencing the process from design, improvement to final implementation of the system.This competition, as a national top competition, gathered outstanding college students from universities across the country, allowing us to compete with future talent in the integrated circuit industry, greatly enhancing our vision and capabilities. We are very grateful to the staff and teachers of the competition organizing committee for their efficient organization and guidance, allowing us to participate in such a competition to showcase our talents and engage with current cutting-edge technologies and design trends.

Future Work

1. This system will further enhance the recognition capability of license plates under rotation, distortion, and other conditions by incorporating more license plate recognition decision-making methods;

2. Adjusting binarization and character segmentation methods to achieve recognition of new energy vehicle license plates;

3. Enhancing the system’s connectivity to make it more applicable in indoor parking management, intelligent parking poles, and other application scenarios.

End

The content of this work comes from the WNS<0 team and is not open source.

About the Arm Education Program

The Arm Education Program connects higher education institutions with a wealth of Arm products, providing educators, researchers, university students, and even middle and primary school students with teaching materials, hardware platforms, software development tools, IP, and resources, supporting the use of Arm technology for educational purposes, and committed to promoting educational reform and cultivating talent in the integrated circuit industry through the Arm innovation ecosystem. The Arm Education Program regularly holds technical seminars, teacher training, shares excellent courses, and conducts ASC practical training, providing a platform for educators and students to communicate and helping them master advanced technologies based on Arm.