Smart traffic light system with automated violation detection | Iot và ứng dụng | Học viện Công nghệ Bưu chính Viễn thông
Red-light enforcement cameras are widely applied technological solutions to enhance urban traffic management efficiency. This paper presents the experimental study, results, and conclusions of a traffic light model integrated with violation detection cameras based on image processing and microcontroller control. The system is designed to detect vehicles running red lights, capture license plate numbers, and store violation information in a database. Tài liệu được sưu tầm và soạn thảo dưới dạng file PDF để gửi tới các bạn cùng tham khảo, ôn tập đầy đủ kiến thức, chuẩn bị cho các buổi học thật tốt. Mời bạn đọc đón xem!
Môn: IoT và ứng dụng 143 tài liệu
Trường: Học viện Công Nghệ Bưu Chính Viễn Thông 1.8 K tài liệu
Tác giả:
Preview text:
The 3 nd
Thematic Workshop on IoT Solutions for Smart Cities (SCIOT3-2025)
IOT-BASED SMART TRAFFIC LIGHT SYSTEM WITH AUTOMATED
VIOLATION DETECTION
Nguyễn Duy Hải Đăng1*, Hoàng Đình Phong2*, Nguyễn Hồng Anh Quân3*, Bùi Hoàng Sơn4*,
Bùi Xuân Trường5
1Department/Unit: Information Technology 1, School/Institute: Posts and Telecommunications Institute of
Technology, Address: Km10, Nguyen Trai Road, Ha Dong District, Hanoi, Vietnam
2Department/Unit: Information Technology 1, School/Institute: Posts and Telecommunications Institute of
Technology, Address: Km10, Nguyen Trai Road, Ha Dong District, Hanoi, Vietnam
3Department/Unit: Information Technology 1, School/Institute: Posts and Telecommunications Institute of
Technology, Address: Km10, Nguyen Trai Road, Ha Dong District, Hanoi, Vietnam
4Department/Unit: Information Technology 1, School/Institute: Posts and Telecommunications Institute of
Technology, Address: Km10, Nguyen Trai Road, Ha Dong District, Hanoi, Vietnam
5Department/Unit: Information Technology 1, School/Institute: Posts and Telecommunications Institute of
Technology, Address: Km10, Nguyen Trai Road, Ha Dong District, Hanoi, Vietnam
*Email: QuanNHA.B22CN662@stu.ptit.edu.vn
Red-light enforcement cameras are widely applied technological solutions to enhance urban traffic management
efficiency. This paper presents the experimental study, results, and conclusions of a traffic light model integrated
with violation detection cameras based on image processing and microcontroller control. The system is designed
to detect vehicles running red lights, capture license plate numbers, and store violation information in a database.
The model is implemented as a scaled-down simulation, allowing for testing algorithms, processing capabilities,
and technical procedures before real-world deployment. Experimental results show that the system operates
reliably, achieving a recognition accuracy of 92% under indoor lighting conditions. The study also proposes
potential improvements, including the integration of artificial intelligence (AI) for enhanced detection, Internet of
Things (IoT) connectivity for real-time monitoring, and synchronization with the national database to manage
violation records comprehensively.
This model holds significant value for education and research, providing students and researchers with
opportunities to understand traffic violation detection algorithms, microcontroller control, and image processing in
a controlled environment. Moreover, insights from this model can inform the development of intelligent traffic
systems in Vietnam, aiming for safer and more efficient urban traffic management. Overall, this study
demonstrates the feasibility and effectiveness of combining traffic light systems with camera-based enforcement
technology, while highlighting the potential for further technological integration, optimization, and application in future smart traffic systems.
Keywords: Red-light camera, traffic management, image processing, microcontroller, intelligent traffic system INTRODUCTION
deliberately running red lights, collect evidence in the
form of license plate images, and store the violation
Effective urban traffic management is a global information in a database. The model is built as a
challenge , especially in rapidly urbanizing cities like scaled-down simulation to provide a controlled testing
those in Vietnam. Traffic law violations, particularly platform. This allows the research team to evaluate
running red lights, are not only a major cause ofthe reliability of image processing algorithms, the
traffic accidents but also significantly reduce traffic processing capacity of the hardware, and the technical
flow and the efficiency of the traffic system. To operating procedures before considering real-world
address this issue, the application of advanced deployment. Experimental results in an indoor
technological solutions has become urgent. One of the lighting environment show that the system operates
most widely applied solutions is the deployment of stably and achieves an impressive license plate
traffic enforcement camera systems at signalized recognition accuracy of 92%. In addition to
intersections. This study introduces and details the demonstrating the feasibility of the current model, this
project, "IoT-Based Smart Traffic Light System with study also proposes potential directions for
Automated Violation Detection". The project
improvement. Future development directions include
combines the traditional traffic light model with integrating Artificial Intelligence (AI) to enhance
advanced image processing technology and a detection accuracy, utilizing Internet of Things (IoT)
microcontroller control system. The main objective of connectivity for real-time monitoring, and
the system is to automatically detect vehicles synchronizing violation data with the national 1 The 3 nd
Thematic Workshop on IoT Solutions for Smart Cities (SCIOT3-2025)
database for comprehensive record management.
Experimental Intersection and Operating
Finally, this model not only has practical value in Conditions
improving traffic safety but also plays a significant
role in education and research. It provides a valuable The traffic light prototype was programmed
practical environment for students and researchers to according to standard signal timing cycles commonly
grasp the principles of violation detection algorithms, applied at urban intersections. The light cycles were
microcontroller control, and image processing,
aligned with the internal clock of the MCU and
thereby contributing to the development of intelligent synchronized with the camera trigger to ensure traffic systems in Vietnam.
accurate correlation between traffic signal state and recorded images. EXPERIMENTAL
Model vehicles were maneuvered through a scaled
Experimental Setup and Evaluation
intersection under multiple predefined scenarios:
To evaluate the performance, reliability, and
Scenario 1: A vehicle purposely runs the
feasibility of the proposed IoT-based smart traffic
red light immediately after the signal turns
light system with automated red-light violation red.
detection, a scaled-down experimental model was
Scenario 2: A vehicle follows traffic
constructed. This simulation-based approach
regulations—slowing down and stopping
provides a controlled environment that allows the
correctly before the stop line.
research team to thoroughly test sensing algorithms,
Scenario 3: A vehicle moves through the
hardware responsiveness, and the overall
intersection under low-light conditions to
coordination between system components prior to evaluate robustness.
real-world deployment. The experimental setup
Scenario 4: A fast-moving vehicle crosses
consists of the following subsystems:
the stop line during the transition from yellow to red.
4.1 System Architecture and Components
Scenario 5: A partial occlusion scenario
where the license plate is partially blocked or angled.
The prototype is composed of three essential modules:
These scenarios simulate the most common real-
world red-light violations and provide a
Microcontroller Unit (MCU):
comprehensive basis for evaluating algorithmic and
The MCU is responsible for regulating the traffic hardware performance.
light cycles (red–yellow–green) and receiving signals
from the camera module. It synchronizes timing data,
communicates with the image-processing subsystem, Violation Detection Mechanism
and triggers event recording logic when potential violations occur.
The detection pipeline involves three major steps:
Red-Light Camera Module: 1. Signal Monitoring:
A camera is positioned facing the stop line of the
The MCU continuously monitors the current
scaled-down intersection to continuously monitor the
traffic light state. Whenever the light enters
movement of vehicles. The camera streams real-time
the red phase, the system automatically
frames to the processing unit, where an image
activates violation detection mode.
processing algorithm is executed to detect license
2. Image Processing and License Plate
plates and determine whether a vehicle has crossed Recognition:
the stop line during the red phase.
The camera transmits image frames to the
processing unit, where a license plate Violation Database:
extraction and recognition algorithm is
All detected violations—including license plate
applied. The algorithm identifies the plate
number, timestamp, vehicle movement state, and
region, segments characters, and classifies
corresponding image evidence—are stored in a
them using pattern-matching techniques.
database. This database enables subsequent retrieval,
3. Violation Determination:
statistical analysis, and evaluation of system
If a vehicle crosses the stop line while the effectiveness.
system is in red-light mode, a violation record is generated. Each record includes: o License plate number 2 The 3 nd
Thematic Workshop on IoT Solutions for Smart Cities (SCIOT3-2025) o Timestamp of violation
These findings highlight the need for improved o Image evidence
algorithms, such as AI-based image processing, noise o
System status (signal state, vehicle
reduction, and hardware enhancements. position)
Educational and Research Value
The data is then uploaded to the database for storage and retrieval.
In addition to technical testing, the prototype serves
as a cost-effective educational tool. Students and Evaluation Metrics researchers can:
To quantify system performance, several key
observe the violation detection pipeline, parameters were measured:
test and compare image processing algorithms, Response Time:
experiment with traffic control logic,
The time interval between the moment a
assess MCU–camera communication
vehicle crosses the red light and the moment reliability, and
the violation is digitally logged in the
explore IoT-based integration scenarios. system.
License Plate Recognition Accuracy:
The safe and controlled environment makes this
Evaluates the correctness of extracted and
model particularly useful for academic research and
identified plate numbers under different
practical training in intelligent transportation
environmental and motion conditions. systems.
Data Storage and Retrieval Efficiency:
Measures the reliability and consistency of
Summary of Experimental Findings
storing multiple violation events in the
database across prolonged operation.
Overall, the experiment validates the feasibility, System Stability:
stability, and effectiveness of the IoT-based smart
Assessed through repeated long-duration
traffic light system with automated violation
testing cycles to evaluate the resilience of
detection. The prototype provides important
the hardware and software integration.
empirical evidence, a reliable test platform, and a
robust dataset for future system enhancements and Experimental Results real-world implementation.
The system demonstrated consistent and stable
RESULT AND DISCUSSION
operation across all test scenarios. Under controlled
indoor lighting conditions, the prototype achieved a
A series of controlled experiments were conducted
license plate recognition accuracy of 92%,
using the scaled-down traffic light model to evaluate
confirming the feasibility of implementing the
the performance of the IoT-based violation detection
recognition algorithm on a low-resource embedded
system. More than 200 test scenarios were executed environment.
under different lighting conditions, vehicle speeds,
and violation behaviors. The evaluation focused on
The response time—from violation occurrence to
four key performance indicators: license plate
data recording—remained within acceptable
recognition accuracy, violation detection
thresholds for real-time monitoring applications.
reliability, response time, and database recording
Data storage tests confirmed that the database was stability.
able to handle repeated entries without loss or corruption.
The system demonstrated stable and consistent
performance throughout the experiment. Under
However, the experiment also revealed limitations:
standard indoor lighting conditions, the license plate
recognition module achieved an average accuracy of
Reduced recognition accuracy in low-light
92%, successfully identifying most vehicles that environments
intentionally ran the red light. The microcontroller–
Difficulty detecting plates when partially
camera synchronization ensured that images were obscured
captured precisely at the time of violation, reducing
Decreased performance when vehicles move the risk of missed detections. at higher speeds
Sensitivity to camera angle and glare
However, the recognition accuracy varied under different conditions: 3 The 3 nd
Thematic Workshop on IoT Solutions for Smart Cities (SCIOT3-2025)
Low-light scenarios reduced accuracy to
These limitations highlight areas where improvement
approximately 85–87% due to noise and
is essential before real-world deployment. insufficient contrast.
Fast-moving vehicles occasionally
The findings show that the system meets the
produced motion blur, which slightly
fundamental requirements of a red-light violation hindered character extraction.
detection framework. However, further
Partially obscured license plates (e.g.,
improvements are necessary to increase robustness,
covered by model parts) resulted in including:
misclassification or unrecognized characters.
1. Integrating AI-based Vision Models (e.g.,
YOLO, OCR Deep Learning) to improve
Despite these limitations, the system consistently
license plate recognition in complex
detected the violation event itself with near-perfect environments.
reliability, even when license plate text extraction
2. Utilizing IoT cloud platforms for real-time was imperfect.
violation monitoring and remote system management.
The average response time—from the moment a
3. Synchronizing data with national traffic
vehicle crossed the stop-line during a red signal to
violation databases, enabling automatic
the moment violation data was fully stored in the ticket generation.
database—was measured at approximately 1.2
4. Upgrading hardware for outdoor use (IR
seconds. This near real-time performance indicates
cameras, higher resolution modules, weather
that the microcontroller, camera module, and image protection).
processing algorithms operate efficiently under the designed hardware constraints.
Beyond technical effectiveness, the model offers
significant educational value. Students and
The stability of this response time remained
researchers can observe the full workflow—from
consistent across multiple test iterations,
image processing to database storage—allowing
demonstrating that the system can handle continuous them to experiment with algorithms, modify
operation without delays or buffering issues.
microcontroller logic, and study IoT-based traffic
management systems in a safe and low-cost
Each detected violation generated a complete data environment. entry consisting of:
Overall, the experimental results demonstrate that the Captured license plate image
IoT-based smart traffic light system is feasible, Extracted license plate number
stable, and effective in detecting red-light violations Timestamp of the violation
within controlled environments. The insights gained
Type of violation (e.g., “Red Light
from this model lay the groundwork for future Running”)
implementation in real-world traffic systems and
Experimental environmental condition
support ongoing research toward smarter and safer urban transportation solutions
The database showed high stability, with no data
loss or corruption across 200+ entries. Retrieval
queries and statistical reporting (e.g., number of
violations per session) were executed quickly,
supporting the feasibility of real-time monitoring applications.
Picture 1: The Traffic Light System Model
Although the experimental results are promising, CONCLUSION
several technical limitations were identified:
This paper has presented a comprehensive study
Reduced recognition accuracy under poor lighting
on the design, implementation, and experimental
Sensitivity to camera angle and plate
evaluation of a scaled-down traffic light model visibility
integrated with a red-light violation detection
Limited robustness to motion blur
camera. The system combines microcontroller-
Indoor testing conditions do not fully reflect based traffic light control with advanced image outdoor environments
processing techniques to automatically detect
and record traffic violations. The experimental 4 The 3 nd
Thematic Workshop on IoT Solutions for Smart Cities (SCIOT3-2025)
results demonstrate that the system operates
contribute significantly to improving traffic
reliably and stably under controlled indoor
safety, reducing violations, enhancing traffic
conditions, achieving an impressive license plate flow, and supporting the development of smart
recognition accuracy of 92%. In addition, the
traffic systems in Vietnam and similar urban
system exhibits fast response times,
contexts. Future work may involve scaling up
approximately 1.2 seconds from the moment a the system for real-world intersections,
vehicle runs a red light to the recording of
integrating with national traffic management
violation data in the database, and effectively
platforms, and exploring advanced AI
stores violation information, including license
algorithms to handle challenging conditions
plate numbers, timestamps, and the violation
such as varying lighting, weather, and occluded
context, enabling efficient retrieval, statistical
license plates. Ultimately, the study confirms
analysis, and management of traffic violations.
that intelligent traffic management systems
combining IoT, AI, and automated violation
The study also identifies certain limitations and detection hold substantial promise for modern
challenges that need to be addressed for real- urban mobility and safety.
world deployment. Recognition accuracy
decreases under low-light conditions, when Acknowledgement
license plates are partially obscured, or when
vehicles move at higher speeds. These factors
This work was carried out with the support
highlight the necessity of further enhancing the regarding suggestions for the research topic idea
image processing and recognition modules,
and valuable research direction from Mr. Đặng
particularly through the integration of artificial Văn Hiếu
intelligence (AI) techniques such as deep
learning-based Convolutional Recurrent Neural References
Networks (CRNN) and attention mechanisms.
Moreover, extending system connectivity via the 1. MDPI – Passable: An Intelligent Traffic Light
Internet of Things (IoT) would enable real-time System with Integrated Incident Detection and
monitoring, remote control, and synchronization
Vehicle Alerting. Sensors.
of violation data with national traffic databases,
creating a more robust and scalable solution for 2. Phand, A., Bagade, S., Bandgar, N., & Wayal, urban traffic management.
G. (2024). Real-Time Traffic Light Optimization
Using AI and IoT. IJRASET.
Beyond technical performance, this research
emphasizes the educational and research value
3. Oloriegbe, D. U., & Ikharo, B. A. (2025).
of the proposed model. The scaled-down system Development of an IoT-Based Traffic Signal
provides students and researchers with a
Violation Detection System. Journal of
practical and safe platform to study
Engineering Research and Reports, 27(2), 1–9.
microcontroller-based traffic control, image
processing algorithms, automated violation
detection, and database management. By
4. Thao, L. Q., Duong, D. C., Nguyen, T. A.,
allowing direct observation of the detection
Pham, M. A., Ha, M. D., & Nguyen, M. (2022).
process and testing under multiple scenarios, the Automatic Traffic Red-Light Violation Detection
model fosters hands-on learning and encourages Using AI. IIETA.
the development of innovative solutions in intelligent traffic systems.
5. Dang, L. T. A., … (2024). Vietnam Vehicle
Number Recognition Based on an Improved
Overall, this study validates the feasibility,
CRNN with Attention Mechanism.
effectiveness, and potential impact of integrating
traffic light control systems with automated 6. Li, Z.,
… (2025). A method for license plate
violation detection cameras. It not only provides recognition in low-light conditions based on
a foundation for future improvements, such as
deep learning. ITM Conferences.
AI-enhanced recognition and IoT-enabled real-
time monitoring, but also demonstrates a clear
7. Lê Thị Thu Hằng (2016). Nghiên cứu về mạng
pathway toward practical deployment in urban
neural tích chập và ứng dụng cho bài toán nhận
traffic environments. The proposed model can
dạng biển số xe. 5 The 3 nd
Thematic Workshop on IoT Solutions for Smart Cities (SCIOT3-2025)
8. Viblo – “Nhận diện text trong hình ảnh với CRNN + CTC”. 6
Tài liệu liên quan:
-
Tóm tắt lý thuyết môn IoT và ứng dụng | Học viện Công Nghệ Bưu Chính Viễn Thông
23 12 -
Accelerated Corner-Detector Algorithms: GPU Implementations and Results | Iot và ứng dụng | Học viện Công nghệ Bưu chính Viễn thông
46 23 -
Đề xuất Dự án IoT: Thiết bị phát hiện chạm cho xe máy | Iot và ứng dụng | Học viện Công nghệ Bưu chính Viễn thông
35 18 -
Phân Tích và Giải Pháp FastAPI | Iot và ứng dụng | Học viện Công nghệ Bưu chính Viễn thông
30 15 -
Tăng cường pháp chế xã hội chủ nghĩa trong quản lý nhà nước | Iot và ứng dụng | Học viện Công nghệ Bưu chính Viễn thông
25 13