Design of a communication system
Images for identification of vehicle plates
Fabrizio Andre Farfán Prado¹, William César Pérez Campos¹, Steisy Anahi Carreño Tacuri¹✉, Favio David Cabrera Alva¹
¹ Facultad de Ingeniería Electrónica y Eléctrica, Universidad Nacional Mayor de San Marcos, Lima, Perú,
Abstract
This work presents the design and implementation of a low-energy wireless image transmission system for vehicle plate
recognition, using the ESP32-CAM and LoRa DXLR01 modules. The system captures images in real time, processes them locally
and transmits them via UART2 to a second ESP32. Subsequently, the data is sent through the LoRa link and stored on the
ThingSpeak platform for remote monitoring. The experimental results show that the system achieves a recognition rate of 92.4%
under optimal lighting conditions and an average transmission latency of 3.2 seconds. The energy efficiency of the system makes
it suitable for applications in access control, vehicular surveillance and infrastructure monitoring
Keywords: License plate recognition, ESP32-CAM, LoRa DXLR01, wireless transmission, ThingSpeak, UART2, low-power
consumption.
1. Introduction
Plate recognition is a key technology in the automation of
access control, surveillance and traffic management. However,
its implementation in environments with limited connectivity
and energy constraints presents several challenges, as
traditional solutions rely on wi-fi networks, cell phones or high
consumption hardware. In this context, lora (long range)
emerges as an efficient technology for wireless transmission,
allowing long range communication with significantly lower
power consumption.
Figure 1: Outline of the capture system
This work proposes a vehicle identification system based on
esp32-cam and lora dxlr01, which captures images of vehicle Other studies have explored the possibility of integrating cloud
plates and transmits them to a receiver node using uart2. The processing to improve number plate recognition. Ashari et al.
data is then stored on the thingspeak platform for real-time (2022) proposed a system based on ESP32-CAM and Amazon
remote viewing. The system is designed to operate in Web Services (AWS), optimising vehicle identification in car
environments where access to high-speed networks is limited, parks. However, the need for network infrastructure for image
offering a low-cost, energy-efficient solution. processing can be a limitation in areas where Internet access is
unstable. To mitigate this problem, some work has evaluated
2. Review of the Literature the use of LoRa as an alternative for real-time image
transmission. Hahn et al. (2023) designed an insect trap
Digital signal processing (PDS) applied to vehicle plate monitoring system using ESP32-CAM and LoRa, highlighting
recognition has been widely investigated in the last decade, the efficiency of this technology for image transmission in low
especially with the integration of artificial vision, low power connectivity environments. Their implementation demonstrated
communication networks and iot devices. The combination of that LoRa can be a viable solution for real-time image
esp32-cam, lora transmission and image recognition transmission without relying on broadband networks.
algorithms has enabled the development of efficient solutions
for real-time data capture and analysis. Kusuma et al. (2023)
developed an intelligent parking system based on esp32-cam to
capture vehicle images and provide contactless access. Its
implementation demonstrated the capability of this device for
real-time vehicle identification and integration with remote
monitoring platforms. However, its reliance on wi-fi or 4g
networks may limit its application in low connectivity
environments.
Figure 2: Diagram showing the location of the LoRa module on the trunk of a
tree. (b) Main components of the system, responsible for capturing images in
the trap and sending the signals from the top of the tower.
Similarly, Nguyen et al. (2023) developed an AI-based system 802.11 b/g/n, Bluetooth v4.2 BLE, and multiple interfaces
for digital water meter recognition using LoRa for long- (SPI, I2C, UART, ADC, DAC). It supports up to 16 MB of
distance image transmission. Their research validates the external Flash memory and low-power modes.
applicability of LoRa in embedded systems, supporting its use
in number plate recognition, as it enables efficient image
transmission without high energy consumption. However, its
focus was limited to static images, which differs from the real-
time processing required in vehicular applications. In addition
to hardware and transmission methods, the accuracy of vehicle
plate identification has been the subject of several research
studies. Dixit et al. (2020) implemented a smart parking
system based on machine vision and IoT, using image
processing to detect vehicles and optimise parking space
management. Their implementation employed convolutional
neural networks (CNNs) and detection algorithms to improve
the accuracy of vehicle identification.
Fig. 4 ESP32 Microcontroller
2) ESP32-CAM: Compact camera based on ESP32, with a 2
MP OV2640 module that captures images and videos in VGA
and QVGA resolutions. It offers Wi-Fi and Bluetooth
connectivity to transmit images over the internet.
Figure 3: Activity diagram of the parking system currently implemented.
On the other hand, Kusumawati and Cahyadi (2017) explored
the application of OCR (Optical Character Recognition) and
image processing in the identification of vehicle number
plates, obtaining high accuracy rates in controlled
environments. However, their performance was affected by
adverse environmental conditions, such as poor illumination or
inadequate capture angles. These limitations highlight the
importance of combining advanced machine vision algorithms
with optimised transmission strategies to improve the
reliability of vehicle monitoring systems.
Comparing previous approaches, it is observed that previous
studies have limitations in terms of network infrastructure Figure 5: ESP32-CAM Microcontroller
dependency, identification accuracy and data transmission
efficiency. The present research seeks to address these
limitations through a solution that integrates ESP32-CAM for 3) LCD 1602: Display module that displays text in a 2-line, 16-
image capture, LoRa for efficient transmission and OCR with character matrix. It uses an HD44780 controller and allows
OpenCV for real-time number plate recognition. This parallel communication. It operates at 5 V and has adjustable
combination will allow the development of a robust and LED backlighting.
autonomous system, without dependence on Wi-Fi networks or
cloud processing, optimising vehicle identification in urban
and rural environments.
3. Metodology
3.1 System Architecture
Figure 6: LCD 1602 display
Components:
1) ESP32: Microcontroller for high-performance IoT
applications, with a 240 MHz dual-core processor, Wi-Fi
4) DLX-LR01: The LoRa DXLR01 is a communication information sent. Then, the ESP32 connected to an LCD
module based on LoRa (Long Range) technology that allows display processes the received data and displays it to the user in
data to be transmitted over long distances with low power a readable format. For transmission to the cloud platform, the
consumption. It is designed for IoT (Internet of Things) ESP32 uses WiFi connectivity to send the data to ThingSpeak.
applications, providing efficient communication in This allows the information to be viewed and analysed
environments where extended range is required. The module is remotely. This design ensures an optimised workflow for real-
ideal for remote monitoring systems, device control and time image-based data transmission using low-power, long-
applications in rural or industrial areas. Its ability to operate at range LoRa communication.
sub-GHz frequencies allows for better penetration in
environments with obstacles, while its low power consumption
makes it suitable for battery-powered devices.
Figure 9: Physical Architecture of the System
B. LoRa communication
The image transmission system for number plate identification
uses LoRa DX-LR01 modules configured to operate in the 433
Figure 7: DLX-LR01 MHz band. Communication between the ESP32 and the LoRa
module is established via UART2, using GPIO pins 12 (RX2)
and GPIO 13 (TX2) with a baud rate of 9600 baud. The DX-
LR01, based on the ASR6601 chipset, employs LoRa (Chirp
3.2 Desarrollo Spread Spectrum) modulation to ensure robust, long-range
communication. The transmitter node captures the image of the
A. Diagrama del Sistema vehicle plate with the ESP32-CAM, extracts the number plate
information and generates a text string in the format
"Plate:ABC123, Link:www.ejemplo.com/imagen.jpg". This
information is transmitted to the LoRa module via AT
commands.
At the receiving node, another ESP32 receives the data via
UART2, extracts the board and link information, and displays
it on an I2C LCD. The data is then sent to ThingSpeak for
storage in the cloud. LoRa communication is configured with
the following parameters: node address (AT+ADDRESS),
network ID (AT+NETWORKID) and operating frequency
Figure 8: System Diagram (AT+BAND=433000000). The DX-LR01 module has an
adjustable transmit power from 0 to +22 dBm and a sensitivity
of up to -138 dBm, allowing a range of up to 6 km in open field
and 3.8 km in urban environments. In addition, its low power
The system proposed in Fig. 8 uses an ESP32-CAM module consumption of 57.78 mA transmit and 6.44 mA receive makes
for image capture and processing, together with a LoRa it suitable for remote monitoring applications. The antenna
module for data transmission. When an image is captured, it used is helical, with an impedance of 50 Ω, optimised for the
undergoes pre-processing, including cropping, brightness 433 MHz band. In terms of transmission speed, the DX-LR01
adjustment and thresholding. Subsequently, the processed module allows configurable data rates depending on the ratio
image is passed to an OCR algorithm implemented in the between coverage and energy efficiency. For this system, the
ESP32-CAM, which extracts the relevant information from the LoRa transmission rate was set at 4.8 kbps, a value that
text. The obtained data is transmitted via the LoRa module provides a balance between range and reliability in urban
using the UART interface for serial communication. environments. However, the module allows higher speeds at
During transmission, the LoRa module modulates the data the sacrifice of range, or lower speeds to improve signal
based on the Chirp Spread Spectrum (CSS) technique, penetration and reduce transmission error rates. To improve
ensuring long-range, low-power communication. The performance, parameters such as Spreading Factor (SF),
communication takes place via the UART of the ESP32, where Bandwidth (BW) and Coding Rate (CR) can be optimised by
the data extracted from the OCR is converted into text strings adjusting the balance between distance and transmission rate
and sent with the LoRaWAN protocol. On the receiving side, according to environmental conditions.
the LoRa module demodulates the signal, retrieving the
C. Image processing with ESP32 CAM Data transmission to the cloud is done via the ESP32, which,
once the message is received via the LoRa module, establishes
As a control and security measure, the system includes an a Wi-Fi connection and sends the processed information to the
ESP32-CAM strategically positioned at the entrance. This ThingSpeak server via the HTTP protocol. For this, the
camera captures an image of the vehicle's number plate at the WiFiClient library is used in conjunction with the ThingSpeak
time of entry. This capture is essential to keep a detailed record REST API, which allows the values in the configured channels
of the cars entering a car park, which is then sent by the LoRa to be updated.
DLx-LR01 to be received and adapted by another LoRA Dlx-
LR01. Once taken, the image is automatically sent to the The system stores two key pieces of data in ThingSpeak:
server via the Wi-Fi network, where it is stored for possible
later use, such as identification in case of incidents. This log -Board Number: Captured and processed by the ESP32-CAM.
not only improves the security of the system, but also allows a
tidy and accessible history to be kept in the cloud. This ensures - Image Link: Generated after processing the photograph of the
efficient and transparent control of vehicle flow. number plate.
The number plate recognition system implemented in this
smart car park uses a combination of OpenCV and Tesseract to These values are stored in ThingSpeak's field1 and field2
automatically detect and read vehicle number plates. First, the fields, allowing them to be queried and analysed later. The
camera connected to the ESP32 captures images of vehicles information is also displayed on the I2C LCD for immediate
entering or leaving the car park. OpenCV takes care of the local display.
image processing, applying pre-processing techniques such as
greyscale conversion, denoising, and edge detection, to This approach allows the processed data to not only be
identify areas that potentially contain the plate. OpenCV then visualised on the LoRa receiver, but also made available in the
performs image segmentation, focusing on the areas most cloud for real-time remote querying and analysis. Furthermore,
likely to contain the number plate. Once the region of interest by using an IoT platform such as ThingSpeak, the system can
has been located, Tesseract, an OCR (Optical Character be integrated with additional services, such as MATLAB
Recognition) engine, is used to extract the alphanumeric Analytics, to perform advanced processing on the collected
characters from the plate. This process automates the data.
registration of vehicles in and out of the system, facilitating
efficient and accurate parking management. 4. Results
The proposed system has been tested in a controlled
environment to evaluate its performance in capturing,
transmitting and displaying number plate images. Multiple tests
were conducted in different scenarios, including varying
illumination conditions and different distances between the
ESP32-CAM and the target vehicles.
During the tests, the ESP32-CAM captured number plate
images with a resolution of 640x480 pixels. The image was
pre-processed locally and sent to the LoRa module via UART2
using GPIO pins 12 (RX2) and GPIO 13 (TX2) with a baud
rate of 9600 baud. The data was transmitted in the format
Figure 10: Process of Plate Recognition ‘Plate: ABC123, Link: www.ejemplo.com/imagen.jpg’
ensuring correct structuring for storage and display.
The LoRa receiver module, configured with the commands AT
AT+ADDRESS=2, AT+NETWORKID=3 and
AT+BAND=433000000, managed to decode the data and
display it on an I2C LCD screen connected to the ESP32.
Subsequently, the information was sent to ThingSpeak via
HTTPClient, where the board number in ‘field1’ and the image
link in ‘field2`’ were correctly stored, allowing remote access
to the captured data.
Figure 11: Plate processed for recognition
Performance metrics evaluated include:
D. Cloud storage - Plate recognition rate: 92.4% in optimal lighting conditions.
- LoRa transmission latency: 3.2 s on average.
The system implements an architecture in which acquired data - LoRa packet loss: 7.8% over distances greater than 1 km.
is sent to the ThingSpeak platform for storage and
visualisation. ThingSpeak is a cloud service that enables real- - Power consumption: 180 mA in transmission and 120 mA in
time data collection and analysis, facilitating remote standby.
monitoring without the need for additional infrastructure.
remote monitoring and optimising the use of available space
(Rivera Laitano and Castro Cornejo, 2016; Avellana
Doménech, 2014).
B. Electronic Tolls and Traffic Control
In electronic toll systems, this technology eliminates the need
to stop at toll booths, as the ANPR system automatically
detects the number plate and generates the corresponding
charge to the user's account, speeding up vehicle flow and
reducing congestion (AS, 2024). A prominent example of this
Figure 12: Plate Displayed on the LCD Screen application is the Port of Bilbao, where road and rail access is
managed through ANPR, allowing for fast validation of
Comparative analysis with other Wi-Fi or LTE-based solutions authorised vehicles and optimising transport logistics (AS,
suggests that LoRa offers better energy efficiency and 2024). However, despite its efficiency, it still faces challenges
coverage in environments with limited connectivity, although such as payment evasion, a problem observed on Spanish roads
it has limitations in data rate. where approximately 3% of drivers avoid toll collection on
5. Applications stretches without physical toll booths (Huffington Post, 2024).
Adoption of this technology continues to increase, with the
A. Smart Parking ANPR market projected to grow from USD 2.79 billion in
2023 to USD 5.95 billion in 2032, driven by the need to
improve urban mobility and road safety globally (Fortune
The Smart Parking system has proven to be an efficient
Business Insights, 2024).
solution for parking management in urban environments,
reducing vehicle congestion and improving the user
C. Implementation in Rural Areas
experience. Its applications cover a variety of scenarios, from
commercial car parks to public and private infrastructures.
ANPR systems allow the automation of access control,
recording vehicle traffic without the need for human
In urban centres, Smart Parking systems allow drivers to locate
intervention. This has been especially useful on large farms and
available spaces in real time through mobile apps or electronic
estates, where security personnel are limited. In addition, the
panels, reducing search time and pollutant emissions
use of this technology facilitates logistics in the management of
(Chancusig Vinocunga and Sánchez Centeno, 2021; Rivera
agricultural machinery and cargo trucks, improving efficiency
Laitano and Castro Cornejo, 2016). Likewise, in shopping
in daily operations (InstantByte, 2024). Another key benefit is
centres and airports, this technology optimises parking space
its integration with video surveillance systems, allowing real-
management through IoT sensors and smart cameras, enabling
time video and photographic evidence to be captured, which
efficient space management and improving vehicle turnover
enhances security in these areas (ViewParking, 2024).
(Chancusig Vinocunga and Sánchez Centeno, 2021).
Figure 13: LT19 Exit Console. Figure 14: Automated Access Control
In the corporate and industrial sector, smart parking facilities In terms of background, Spain has been a pioneer in the
facilitate vehicle access and control through automatic adoption of ANPR in rural areas. In the municipality of Soto
identification of number plates, automated payment systems del Real, number plate recognition cameras have been installed
and data analysis to improve operational efficiency (Rivera to monitor access and detect unauthorised vehicles, increasing
Laitano and Castro Cornejo, 2016; Avellana Doménech, security and reducing the risk of theft in rural properties
2014). In the residential sector, solutions have been (Cadena SER, 2024). Similarly, in Navarre, the Helpbidea
implemented that allow space reservation and integration with system has been implemented to assign unique identifiers to
security systems, providing greater convenience to residents isolated hamlets and hamlets, facilitating the control and
(Chancusig Vinocunga and Sánchez Centeno, 2021). management of areas that are difficult to access (El País, 2024).
Furthermore, integration with IoT communication networks 6. Discussion
and technologies such as LoRa and ESP32-CAM allows the
implementation of Smart Parking systems in areas that are The results obtained validate the feasibility of the ESP32-CAM
difficult to access or have infrastructure limitations, facilitating and LoRa-based system for real-time vehicle identification.
Compared to solutions relying on Wi-Fi or mobile networks, be an efficient and low-energy solution. Using the ESP32-
the present system offers a more efficient alternative in terms CAM and LoRa DXLR01 modules, reliable vehicle
of energy consumption and coverage in areas without identification was achieved in environments with limited
traditional network infrastructure. connectivity.
The main advantage of LoRa technology is its ability to The system achieved a recognition rate of 92.4% under optimal
transmit data over long distances with low power lighting conditions and an average transmission latency of 3.2
consumption. However, tests revealed that the packet loss rate seconds, making it suitable for real-time vehicle monitoring
increases significantly over distances greater than 1 km, and access control applications. However, it was observed that
suggesting the need to implement error correction techniques in poorly lit scenarios the recognition rate was reduced to
or improve transmission power in dense urban environments. 78.5%, which highlights the need to incorporate image
enhancement techniques or complementary lighting sensors.
Another relevant aspect is the dependence on illumination for
optimal image capture. It was observed that in low-light In terms of data transmission, LoRa proved to be a viable
environments, the plate recognition rate decreased to 78.5%, technology for long-distance information delivery with reduced
which could be addressed by image enhancement techniques energy consumption. However, a packet loss of 7.8% was
or the use of infrared light sensors. identified at distances greater than 1 km, suggesting future
implementation of error correction techniques or transmission
Finally, the integration with ThingSpeak allowed a secure and power adjustment. Integration with the ThingSpeak platform
accessible storage of data via REST API, making it easy to enabled remote data storage and retrieval, facilitating real-time
query in real time from any device with an internet connection. monitoring without relying on traditional network
This approach highlights the potential of the system for infrastructure. This positions the system as an efficient
applications in access control, vehicle monitoring and alternative for applications in urban and rural environments
intelligent parking management, with the possibility of with limited connectivity.
expanding to other sectors requiring real-time image
recognition and efficient transmission via LoRa. As future lines of research, the optimization of the plate
recognition algorithm to improve its performance in adverse
conditions is proposed, as well as the exploration of image
7. Conclusions
compression methods to reduce transmission latency. In
addition, the integration of artificial intelligence could improve
The design and implementation of the wireless image
vehicle recognition accuracy, broadening the scope of
transmission system for vehicle plate recognition has proven to
application of the proposed system.
References
AS. (2024). Un puerto español se suma a la vanguardia tecnológica con
el super software para controlar accesos. AS. Recuperado de
https://as.com
Ashari, I.F., Satria, M.D. and Idris, M., 2022. Parking system
optimization based on IoT using face and vehicle plate recognition via
Amazon Web Service and ESP32-CAM. Computer Engineering and
Applications, 11(2), pp.137-152.
Avellana Doménech, E. (2014). Ciudad Inteligente (Smart City), Gandía.
Propuestas para un plan de actuación en el sector turístico.
Universidad Politécnica de Valencia.
Cadena SER. (2024). Soto del Real contará con cámaras OCR para leer
matrículas de vehículos en tiempo real. Recuperado de
https://cadenaser.com
Chancusig Vinocunga, D. J., & Sánchez Centeno, J. D. (2021). Estudio de
factibilidad para la concesión del servicio de parqueadero del
Aeropuerto Internacional Cotopaxi, Provincia de Cotopaxi. Escuela
Superior Politécnica de Chimborazo.
Dixit, M., Srimathi, C., Doss, R. and Loke, S., 2020. Smart Parking with
Computer Vision and IoT Technology. IEEE TSP 2020, pp.170-172.
El País. (2024). Los caseríos navarros ya tienen DNI y plano de metro
para evitar desgracias. Recuperado de https://elpais.com
Fortune Business Insights. (2024). Mercado de sistemas ANPR:
Crecimiento y tendencias hasta 2032. Fortune Business Insights.
Recuperado de https://fortunebusinessinsights.com
Hahn, F., Valle, S., Rendón, R., Oyorzabal, O. and Astudillo, A., 2023.
Mango Fruit Fly Trap Detection Using Different Wireless
Communications. Agronomy, 13(1736), pp.1-24.
Huffington Post. (2024). El “escaqueo” masivo de conductores en el pago
de peajes en España. Huffington Post. Recuperado de
https://huffingtonpost.es
InstantByte. (2024). Sistema para el reconocimiento de matrículas de
Dahua ANPR. Recuperado de https://www.instantbyte.com
Kusuma, V.A., Arof, H., Suprapto, S.S., Suharto, B., Sinulingga, R.A.
and Ama, F., 2023. An internet of things-based touchless parking
system using ESP32-CAM. International Journal of Reconfigurable
and Embedded Systems, 12(3), pp.329-335.
Kusumawati, R. and Cahyadi, W.A., 2017. Implementation of OCR-
based vehicle license plate recognition using image processing.
International Journal of Electrical and Computer Engineering, 7(1),
pp.45-52.
Nguyen, T.T., Pham, H.V. and Le, D.T., 2023. Implementing AI on low-
power embedded devices for digital water meter identification and
data transfer via LoRa network. Journal of Science and Technology,
16(3), pp.45-56
Rivera Laitano, O. D., & Castro Cornejo, J. U. (2016). Estudio de
prefactibilidad para la implementación de un estacionamiento vertical
automatizado en el Centro Histórico de Tegucigalpa. Universidad
Tecnológica Centroamericana (UNITEC).
ViewParking. (2024). Ventajas de utilizar un sistema de control de
accesos con tecnología de reconocimiento de matrículas. Recuperado
de https://viewparking.net
