AI-driven UAV system for autonomous vehicle tracking and license plate recognition

Ahad Alotaibi; Chris Chatwin; Phil Birch

doi:10.1515/eng-2024-0101

Article Open Access

AI-driven UAV system for autonomous vehicle tracking and license plate recognition

Ahad Alotaibi , Chris Chatwin and Phil Birch

Published/Copyright: February 22, 2025

Published by

Become an author with De Gruyter Brill

Submit Manuscript Author Information

From the journal Open Engineering Volume 15 Issue 1

Abstract

The integration of Artificial Intelligence (AI) with image processing and autonomous flight capabilities in Unmanned Aerial Vehicles (UAVs) represents a significant advancement in modern surveillance and tracking systems. This research explores a novel method for locating vehicles with pre-identified license plate numbers through an AI-enhanced framework. The proposed system captures vehicle plate details and stores them for subsequent comparison. Autonomous UAVs are deployed within a predefined area to capture high-resolution images of vehicle plates, which are then processed and analysed using advanced AI algorithms designed for optical character recognition and machine learning. Recognized plate numbers are matched against pre-stored entries in real-time. Upon identification of a match, the system accurately determines and displays the vehicle’s location, providing precise geospatial data. This approach demonstrates high precision and efficiency in vehicle tracking, significantly improving upon conventional surveillance techniques, which often rely on manual monitoring and static camera setups. The AI-driven system not only enhances the accuracy of vehicle identification but also reduces the time and human resources required. This study also explores the broader implications and potential applications of this advanced tracking system across various sectors. In law enforcement, it enables real-time tracking of stolen vehicles or suspects. In traffic management, it assists in monitoring and managing vehicle flow and enforcing parking regulations. In security monitoring, it enhances perimeter security by identifying unauthorized vehicles in restricted areas. This research underscores the system’s robustness and adaptability in various practical applications, marking a significant step forward in the field of automated surveillance and vehicle tracking.

Keywords: AI; image processing; UAVs; OCR; systems integration

1 Introduction

The rapid evolution of technology in recent decades has seen Artificial Intelligence (AI) emerge as a transformative force across various domains. Among its numerous applications, the integration of AI with Unmanned Aerial Vehicles (UAVs) has garnered significant attention, particularly in the fields of surveillance and tracking. UAVs, or drones, equipped with AI capabilities, offer a versatile platform for tasks that require real-time data collection, analysis, and decision-making. This synergy of AI and UAV technology has the potential to revolutionize traditional surveillance methods, providing more efficient and accurate solutions. One of the most promising applications of AI-driven UAVs is in the domain of vehicle tracking and license plate recognition (LPR). Conventional vehicle tracking systems often rely on fixed surveillance cameras and manual monitoring, which can be labour-intensive and limited in coverage [1]. These systems are particularly challenged in dynamic environments where vehicles are constantly on the move, necessitating rapid data processing and real-time response capabilities. The incorporation of AI with UAVs presents a novel solution to these challenges, enabling autonomous vehicle tracking with enhanced precision and efficiency.

This research delves into a pioneering approach to locate vehicles with pre-identified license plate numbers using an AI-enhanced framework. The primary objective of this study is to develop a system that employs UAVs for autonomous vehicle tracking and LPR, thereby addressing critical needs in real-time surveillance and monitoring. The proposed system captures vehicle plate details and stores them for subsequent comparison. Autonomous UAVs are deployed to operate within designated areas, tasked with capturing high-resolution images of vehicle license plates. These images are then processed and analysed using segmentation AI algorithms designed for optical character recognition (OCR) and machine learning, ensuring accurate extraction and recognition of the plate numbers. Upon recognition, the plate numbers are matched against pre-stored entries in real-time. When a match is identified, the system accurately determines and displays the vehicle’s location, providing precise geospatial data. This method demonstrates significant improvements in vehicle tracking accuracy and efficiency compared to traditional surveillance techniques. By reducing the dependency on manual monitoring and static camera setups, the AI-driven system optimizes resource utilization and enhances the overall effectiveness of vehicle identification.

The motivation for this research is rooted in the growing demand for scalable and efficient surveillance solutions across various sectors. In law enforcement, the capability to track stolen vehicles or suspects in real-time can significantly enhance operational efficiency and response times [2]. For traffic management, the system provides tools for monitoring and managing vehicle flow, as well as enforcing parking regulations [3]. In security monitoring, the technology can bolster perimeter security by identifying unauthorized vehicles in restricted areas, thereby preventing potential security breaches [4].

The use of UAVs for surveillance is not new, but their potential has been significantly amplified by the integration of AI technologies [5]. UAVs provide a flexible and mobile platform for surveillance, capable of covering large areas and accessing locations that are difficult or dangerous for humans to reach. AI, on the other hand, offers powerful tools for processing and analysing the vast amounts of data that UAVs can collect. Together, they form a potent combination that can transform traditional surveillance practices. In recent years, AI algorithms have advanced considerably, particularly in the fields of image processing and machine learning [6]. OCR technology, a key component of the proposed system, has seen significant improvements in accuracy and speed. Machine learning algorithms have also become more sophisticated, capable of learning from large datasets and improving their performance over time. These advancements have made it possible to develop systems that can automatically recognize and track vehicles with high accuracy.

The integration of these technologies with UAVs presents several technical challenges. UAVs must be capable of capturing high-quality images of vehicle plates while in motion, which requires advanced stabilization and imaging technology [7]. The AI algorithms used for image processing and recognition must be efficient enough to operate in real-time, enabling the system to quickly identify matches and provide accurate location data. Additionally, the system must be robust enough to operate in various environmental conditions and adapt to different surveillance scenarios.

Traditional methods of vehicle tracking and surveillance are often limited in their effectiveness and efficiency. Static cameras provide limited coverage and require significant infrastructure investment, while manual monitoring is labour-intensive and prone to errors [8]. There is a clear need for a more flexible, accurate, and efficient solution that can operate in dynamic environments and provide real-time data.

The integration of AI and UAV technologies for surveillance and tracking has been the subject of considerable research interest. Several studies have explored different aspects of this integration, each contributing to the body of knowledge in unique ways [9]. In the field of UAV-based surveillance, early research focused on the capabilities of UAVs to capture aerial imagery and monitor large areas. Goodrich et al. [10] investigated the use of UAVs for search and rescue operations, highlighting the potential of UAVs to cover vast terrains quickly and efficiently. This study laid the groundwork for understanding the operational benefits of UAVs in surveillance.

With the advent of AI, researchers began exploring the potential of combining UAVs with AI algorithms to enhance surveillance capabilities. A significant body of work has focused on developing AI algorithms for object detection and tracking. Redmon [11] introduced the You Only Look Once algorithm, which revolutionized real-time object detection by offering high-speed and accurate performance. This algorithm has been widely adopted in various surveillance applications, including vehicle tracking. LPR has also been a major area of research within the broader field of image processing. Anagnostopoulos et al. [12] provided a comprehensive survey on LPR systems, discussing the various techniques used for plate detection, segmentation, and character recognition. Their work highlighted the challenges in LPR, such as variations in plate designs, lighting conditions, and image resolutions.

More recent studies have specifically addressed the integration of LPR with UAVs. For example, Niu et al. [13] developed a UAV-based system for traffic monitoring that uses deep learning algorithms to recognize vehicle license plates from aerial images. Their system demonstrated the feasibility of UAV-based LPR, but also highlighted the need for further improvements in algorithm accuracy and processing speed. Another significant contribution to this field is the work of Ranjan et al. [14], who developed an AI-driven UAV system for automated parking enforcement. Their system employs convolutional neural networks for real-time LPR and integrates with municipal databases to identify parking violations. This study demonstrated the practical applications of AI-UAV systems in urban environments.

Despite these advancements, several challenges remain in the development of fully autonomous AI-UAV systems for vehicle tracking and LPR [15]. These include ensuring high accuracy and reliability under varying environmental conditions, achieving real-time processing capabilities, and integrating with existing surveillance infrastructure.

This research builds upon the existing literature by addressing these challenges and proposing a comprehensive solution for autonomous vehicle tracking and LPR using AI and UAV technologies. By optimizing AI algorithms and leveraging the mobility of UAVs, this study aims to provide a scalable and efficient system that can be applied across various sectors, enhancing the effectiveness of surveillance and tracking operations. The architecture of the proposed system is shown in Figure 1.

Figure 1

Architecture of the proposed system.

The subsequent sections of this article will delve into the methodology, experimental results, and conclusions drawn from this study, providing a comprehensive understanding of the system’s capabilities and future directions for research and further development.

2 Materials and methods

2.1 Inspected car application

The “Inspected Car” application is an innovative tool designed to leverage the capabilities of AI and autonomous UAVs for vehicle tracking and LPR. This application represents a significant advancement in surveillance technology, offering a robust and efficient method for identifying and locating vehicles with pre-identified plate numbers. Developed using modern software tools and AI algorithms, including a segmentation AI algorithm modified for the specific purpose of image processing. the Inspected Car application provides an effective solution for image processing and vehicle tracking. The customized segmentation algorithm is integral to the Plate Reader application, where it is used to accurately process images captured by UAVs, ensuring precise identification and location of vehicles.

The Inspected Car application was developed using the Flutter framework, which utilizes the Dart programming language. Flutter was chosen for its ability to create high-performance, cross-platform applications with a single codebase [16]. The development process involved using Android Studio as the primary Integrated Development Environment (IDE), which facilitated the integration of various development tools and libraries necessary for building a sophisticated application [17].

The Inspected Car application features a user-friendly interface that allows users to input vehicle plate details easily. Users are required to enter the administrative number and the vehicle number, which are specific identifiers for each car. For example, a typical Kuwait car plate includes a two-digit administrative number and a five-digit vehicle number. Once the data are entered, it is stored locally in a SQLite database. SQLite was chosen for its lightweight nature and efficiency in handling local data storage without the need for a remote server [18]. This ensures fast response times and reliable performance even in offline scenarios. Screenshots of the inspected car application are shown in Figure 2.

Figure 2

Screenshots of the inspected car application.

The Inspected Car application is built on a robust architecture that ensures reliability, efficiency, and scalability. The main components of the architecture include the Flutter framework, which allows for the development of a cross-platform application that can run on both Android and iOS devices. Flutter’s rich set of pre-built widgets and fast development cycle make it an ideal choice for building feature-rich mobile applications. SQLite provides a lightweight, serverless database solution that is embedded within the application. This ensures that all data are stored locally, reducing dependency on external servers and improving data access speeds.

The Inspected Car application offers several benefits across various use cases. In law enforcement, the application can be used by agencies to track stolen vehicles or vehicles involved in criminal activities. The ability to capture and analyse license plates in real-time enhances the effectiveness of surveillance operations. Traffic management authorities can use the application to monitor and manage vehicle flow, enforce parking regulations, and identify vehicles that violate traffic rules. In private or restricted areas, the application can be used to identify unauthorized vehicles, thereby enhancing security and preventing potential security breaches. The Inspected Car application represents a significant step forward in the integration of AI and UAV technologies for vehicle tracking and LPR. By leveraging the capabilities of Flutter for cross-platform development, SQLite for efficient data storage, and advanced AI algorithms for OCR, the application provides a robust and scalable solution for real-time surveillance needs. The detailed functionalities and technical architecture ensure that the Inspected Car application is well-equipped to meet the demands of modern surveillance and security operations.

2.2 Inspection mission using Dronelink software

The “Inspected Car” application is an innovative tool designed to leverage the capabilities of AI and autonomous UAVs for vehicle tracking and LPR. A critical component of this system is the inspection mission, which is executed using Dronelink software. This software enables precise control over the UAV’s flight path, ensuring comprehensive coverage of the designated area and effective image capture. This section provides a detailed overview of how Dronelink software is utilized to conduct inspection missions, from planning to execution. An example of an inspection mission is shown in Figure 3.

Figure 3

An example of inspection mission.

Dronelink is a versatile software platform that allows for the creation of complex UAV missions [19]. It is particularly well-suited for inspection tasks due to its ability to automate flight paths and ensure consistent data collection. In the context of the Inspected Car application, Dronelink is used to plan and execute missions that involve capturing high-resolution images of vehicle plates within a specified area [20]. The mission planning process begins with defining the inspection area. This involves specifying the geographical boundaries within which the UAV will operate. Dronelink’s intuitive interface allows users to draw the inspection area on a map, setting waypoints and defining the flight path that the UAV will follow. The software supports various flight modes, including waypoint navigation, grid, and orbit, each offering different advantages depending on the inspection requirements. Once the inspection area is defined, the next step is to optimize the flight path. Dronelink provides advanced tools for optimizing the UAV’s route to ensure maximum coverage and efficiency [21]. The flight path is designed to cover the entire inspection area systematically, minimizing gaps and overlaps. This ensures that all vehicles within the area are captured in the images taken by the UAV. During this stage, parameters such as altitude, speed, and camera angle are set. The altitude is chosen based on the desired image resolution and the UAV’s capabilities. Flying at a lower altitude generally provides higher resolution images, which are crucial for accurate LPR. However, it also requires more passes to cover the same area. The speed is adjusted to balance between coverage time and image quality, ensuring that the images are not blurred. The camera angle is optimized to capture clear and unobstructed views of the vehicle plates. With the flight path and parameters set, the mission is ready to be executed. Dronelink automates the UAV’s flight, guiding it along the predefined path and capturing images at specified intervals. The software continuously monitors the UAV’s status, providing real-time feedback on its position, speed, altitude, and battery life. This allows operators to make adjustments if necessary, ensuring the mission’s success. During the mission, the UAV captures high-resolution images of the surrounding area, including all visible vehicle plates. These images are stored on the UAV’s onboard memory and later transferred to the companion application for analysis. Dronelink’s automation capabilities significantly reduce the workload on human operators, allowing them to focus on monitoring the UAV’s performance and ensuring that the mission’s objectives are met.

After the UAV completes its mission, the captured images are transferred to the Plate Reader application for analysis. This companion application processes the images using advanced AI algorithms to extract and recognize license plate numbers. The recognized plate numbers are then compared with the entries stored in the SQLite database to identify matches. The success of the inspection mission heavily relies on the quality and coverage of the captured images. Dronelink’s precise flight control and automation capabilities ensure that the images are captured at the right angles and intervals, maximizing the chances of accurate plate recognition. By systematically covering the inspection area and capturing high-quality images, the Dronelink software plays a crucial role in the overall effectiveness of the Inspected Car application.

The integration of Dronelink software into the Inspected Car application’s workflow is a key factor in its ability to perform efficient and accurate vehicle tracking and LPR. Dronelink’s advanced mission planning, flight path optimization, and automated execution capabilities ensure comprehensive coverage and high-quality data collection. This, in turn, enhances the performance of the AI algorithms used for image analysis, leading to reliable and accurate vehicle identification. The use of Dronelink software exemplifies the synergy between cutting-edge UAV technology and AI-driven applications, providing a powerful tool for modern surveillance and tracking needs.

2.3 Plate Reader application

The Plate Reader application is an integral component of the Inspected Car system, designed to process and analyse the images captured during UAV inspection missions. Developed using state-of-the-art software tools and advanced AI algorithms, the Plate Reader application ensures accurate recognition and comparison of vehicle license plate numbers. This section provides a comprehensive overview of the Plate Reader application, detailing its development, functionalities, and the critical role it plays in the vehicle tracking process.

Flutter was selected for its capability to build high-performance, cross-platform applications with a single codebase [22]. This choice allows for consistent performance and user experience across both Android and iOS devices. Android Studio was used as the primary IDE for the development of the application, enabling seamless integration of various development tools and libraries.

The application leverages advanced AI algorithms for OCR and machine learning. These algorithms are implemented using TensorFlow Lite, a lightweight version of TensorFlow designed specifically for mobile and embedded devices [23]. TensorFlow Lite enables the application to perform complex image recognition tasks efficiently on the device itself, ensuring real-time processing capabilities [24]. The Plate Reader application performs several key functions essential for the accurate recognition and comparison of vehicle license plate numbers. The primary workflow of the application involves three main stages: image processing, plate recognition, and data comparison as shown in Figure 4.

Figure 4

Workflow of plate reader application.

2.3.1 Image processing

The first stage in the Plate Reader application’s workflow is image processing. This stage is crucial for enhancing the quality of the images captured by the UAV to prepare them for accurate recognition of license plate numbers. Upon completion of the UAV’s inspection mission, the captured images are transferred to the Plate Reader application. These images are typically high-resolution to ensure that details such as license plate numbers can be accurately extracted. Several pre-processing steps are applied to the images to enhance their quality: resizing to a standard dimension to ensure uniformity in processing and to optimize the performance of the OCR algorithm; noise reduction techniques to remove any extraneous data that could interfere with the recognition process, including filtering out grainy textures or irrelevant background details; contrast adjustment to improve the visibility of the license plate numbers by enhancing the difference between the characters and the background; and cropping to isolate the region of interest (ROI), which is the area likely to contain the license plate, reducing the amount of data the OCR algorithm needs to process and making it more efficient [25,26,27]. These pre-processing steps ensure that the images are in the best possible condition for the subsequent recognition phase.

2.3.2 Plate recognition

The second stage of the workflow is plate recognition, where the application uses advanced OCR algorithms to detect and extract the license plate numbers from the processed images. The OCR process begins with detecting the ROI within each image that contains the license plate. This involves analysing the image to locate rectangular shapes and patterns consistent with license plates. Machine learning techniques are used to improve the accuracy of this detection step by training the model on various images of license plates under different conditions [28]. Once the license plate region is identified, the next step is character segmentation. This involves isolating each character on the plate so that they can be individually recognized. Segmentation algorithms are used to detect the boundaries of each character, ensuring they are separated correctly despite variations in spacing and alignment [29]. After segmenting the characters, the OCR algorithm is applied to recognize each character individually. The Plate Reader application uses a pre-trained machine learning model developed with TensorFlow Lite. This model has been trained on a large dataset of license plate images to recognize characters with high accuracy, even under varying conditions such as different fonts, lighting, and angles. The recognized characters are then combined to form the complete license plate number. This digital representation of the license plate is temporarily stored for the next stage.

2.3.3 Data comparison

The final stage of the Plate Reader application’s workflow is data comparison. In this stage, the recognized license plate numbers are compared with the entries stored in the SQLite database to identify matches. The Plate Reader application reads the data from the SQLite database, which contains the details of vehicles with known license plate numbers. The database is stored locally, ensuring fast access and efficient data retrieval. The recognized license plate numbers are compared against the entries in the SQLite database, involving checking each recognized plate number against the stored entries to find an exact match. The comparison algorithm is optimized for speed and accuracy, ensuring that matches are identified quickly. If a match is found, the application retrieves the associated location information (latitude and longitude) from the database. This location data indicates where the vehicle was identified, providing precise geospatial coordinates. The application then notifies the user of the match, displaying the recognized plate number and the vehicle’s location. If no match is found, the application alerts the user that no corresponding record exists in the database.

The Plate Reader application features a user-friendly interface designed to facilitate easy interaction and efficient processing. Users can upload images captured during the inspection mission, initiate the analysis process, and view the results. The interface provides clear visual feedback, displaying the recognized plate numbers and their comparison results. The application also includes functionalities for managing the SQLite database, allowing users to add, update, or delete vehicle records as needed. This flexibility ensures that the database remains up-to-date and relevant for ongoing surveillance operations. Screenshots of the Plate Reader application is shown in Figure 5.

Figure 5

Screenshots of plate reader application.

Once the application successfully reads the license plate number from the uploaded image, it proceeds to compare this number against the database of saved plate numbers. This comparison process is swift and efficient, leveraging advanced matching algorithms to ensure accuracy. When a match is identified, the application retrieves and displays the precise location of the vehicle, as recorded during the UAV’s mission using Dronelink software. This feature provides users with real-time geospatial information, enhancing the system’s utility for tracking and locating vehicles. The ability to accurately identify and report the location of a matched vehicle underscores the effectiveness of the integration between the AI-driven recognition algorithms and the UAV’s detailed data capture capabilities. This scenario is visually depicted in Figure 6, illustrating the seamless interaction between the various components of the system to deliver accurate and timely information to the user.

Figure 6

Screenshots of a registered car in plate reader application.

If a picture of a car is uploaded to the Plate Reader application and it does not match any records in the database, the application will notify the user that the vehicle is not found. This functionality is crucial for maintaining the accuracy and reliability of the system, preventing false positives and ensuring that only verified matches are reported. An example of the user interface displaying this notification for an unregistered vehicle is illustrated in Figure 7. This feature underscores the thoroughness of the system in verifying and reporting vehicle information, thus maintaining a high standard of data integrity and user trust.

Figure 7

Screenshots of an unregistered car in plate reader application.

The Plate Reader application is a vital component of the Inspected Car system, providing the necessary tools for accurate and efficient LPR and comparison. Developed using the Flutter framework and advanced AI algorithms, the application ensures high performance and accuracy in real-world scenarios. By enabling real-time processing and providing a user-friendly interface, the Plate Reader application enhances the overall effectiveness of the Inspected Car system in various applications, including law enforcement, traffic management, and security monitoring. The integration of cutting-edge technologies and meticulous development processes ensures that the Plate Reader application remains a robust and reliable tool for modern surveillance needs.

3 Results

The implementation of the Inspected Car and Plate Reader applications was tested extensively to evaluate their performance in real-world scenarios. The testing involved multiple stages, including data entry, UAV mission execution, image capture, and LPR and comparison. This section presents the detailed results of these tests, highlighting the accuracy, efficiency, and reliability of the system.

3.1 Data entry and storage

The data entry functionality of the Inspected Car application was tested by inputting details of various vehicles, including administrative numbers and vehicle numbers. The SQLite database effectively stored these details locally on the device. Testing showed that the database’s response time for storing and retrieving data was instantaneous, ensuring smooth and uninterrupted user interaction. No significant performance issues were observed, even when the database size increased.

3.2 UAV mission execution

The UAV missions were planned and executed using Dronelink software, which enabled precise control over the UAV’s flight path, ensuring comprehensive coverage of the designated area. During testing, UAVs consistently followed the pre-defined paths with high accuracy, achieving an average deviation of less than 1 m from the planned route. The system was tested under varying environmental conditions, including different lighting and weather scenarios. In over 95% of cases, the UAVs maintained stable flight and produced clear images, even in low-light conditions and moderate wind speeds of up to 20 km/h. The UAVs’ reliability in both image capture and flight stability was quantitatively confirmed through repeated trials, with a success rate of 98% in capturing usable images across all test scenarios

3.3 Image capture and quality

The captured images were analysed for quality and clarity. High-resolution images were obtained, which were crucial for the subsequent OCR process. Image preprocessing steps, including resizing, noise reduction, and contrast adjustment, effectively enhanced the quality of the images. The ROI containing the license plates was clearly identifiable in most images. Some challenges were encountered with images captured under low-light conditions or with significant glare, but these were mitigated through preprocessing techniques.

3.4 LPR

The Plate Reader application utilized advanced OCR algorithms to detect and extract license plate numbers from the processed images. The recognition accuracy was tested across a diverse set of images, including plates with different fonts, sizes, and orientations. The AI model demonstrated high accuracy in recognizing characters, with an overall recognition rate of 95%. Errors primarily occurred in cases where the plates were partially obscured or the characters were highly stylized. However, the system’s ability to handle a wide range of conditions and consistently produce accurate results was commendable.

3.5 Data comparison and match identification

The recognized license plate numbers were compared against the entries stored in the SQLite database. The comparison process was efficient, with matches being identified and displayed in less than 2 s. The system accurately retrieved the location information for matched vehicles, displaying precise geospatial coordinates. In cases where no match was found, the application correctly notified the user, ensuring clear communication of results. Figure 8 shows an example of the results.

Figure 8

Data comparison and match identification.

4 Discussion

The Inspected Car and Plate Reader applications demonstrated robust performance across all testing stages. The integration of AI algorithms and UAV technology proved to be highly effective for vehicle tracking and LPR. The system’s ability to operate in real-time, from data entry to match identification, highlights its potential for practical applications in various sectors. Several challenges were encountered during testing, particularly related to image quality and OCR accuracy. Low-light conditions and glare, which affected image clarity, which in turn impacted recognition accuracy. However, these issues were largely mitigated through preprocessing techniques such as contrast adjustment and noise reduction. Additionally, the use of a diverse training dataset for the AI model helped improve recognition accuracy for license plates with varying characteristics.

The system’s performance indicates its suitability for various applications. In law enforcement, the ability to track stolen vehicles or those involved in criminal activities in real-time can significantly enhance operational efficiency. Traffic management authorities can benefit from improved monitoring and management of vehicle flow, as well as enforcing parking regulations. Security monitoring in restricted areas can also be enhanced by identifying unauthorized vehicles, thereby preventing potential security breaches. Despite the system’s strong performance, there are areas for future improvement. Enhancing the AI model’s capability to handle highly stylized or partially obscured characters can further increase recognition accuracy. Additionally, integrating more advanced image enhancement techniques can improve the system’s robustness under challenging environmental conditions. Expanding the system to support additional data sources and improving its scalability will also be crucial as it is deployed in larger, more complex environments.

5 Conclusion

The Inspected Car and Plate Reader applications represent a significant advancement in the integration of AI and UAV technologies for vehicle tracking and LPR. The system demonstrated high accuracy, efficiency, and reliability in real-world testing scenarios, proving its potential for practical applications in law enforcement, traffic management, and security monitoring. The challenges encountered were effectively mitigated, and the system’s overall performance was robust.

Future improvements will focus on enhancing the system’s capability to handle diverse and challenging conditions, as well as expanding its scalability and data integration capabilities [30,31]. The successful implementation of this system underscores the transformative potential of combining AI and UAV technologies, offering a powerful tool for modern surveillance and tracking needs. The detailed functionalities and robust architecture ensure that the Inspected Car and Plate Reader applications are well-equipped to meet the demands of contemporary surveillance operations, providing a reliable and scalable solution for various sectors

Acknowledgements

The authors acknowledge the allocation of computing resources by The University of Sussex.

Funding information: Authors state no funding involved.
Author contributions: All authors have accepted responsibility for the entire content of this manuscript and consented to its submission to the journal, reviewed all the results and approved the final version of the manuscript. Writing – original draft preparation: Ahad Alotaibi; writing – review and editing: Chris Chatwin; supervision: Chris Chatwin and Phil Birch. All authors have read and agreed to the published version of the manuscript.
Conflict of interest: Authors state no conflict of interest.
Ethical approval: Not applicable.
Consent to participate: Not applicable.
Consent to publication: Not applicable.
Data availability statement: The data presented in this study are available on request from the corresponding author. The data are not publicly available due to privacy.

References

[1] Ibraheem IK, Hadi SW. Design and implementation of a low-cost secure vehicle tracking system. In 2018 International Conference on Engineering Technology and their Applications (IICETA). IEEE; 2018. p. 146–50.10.1109/IICETA.2018.8458096Search in Google Scholar

[2] Sorensen SL. SMART mapping for law enforcement settings: integrating GIS and GPS for dynamic, near-real time applications and analysis. Crime Mapp Crime Prev. 1997;8:349–78.Search in Google Scholar

[3] Aalsalem MY, Khan WZ, Dhabbah KM. An automated vehicle parking monitoring and management system using ANPR cameras. In 2015 17th International Conference on Advanced Communication Technology (ICACT). IEEE; 2015. p. 706–10.10.1109/ICACT.2015.7224887Search in Google Scholar

[4] Cai Z, Liu Z, Kou L. Reliable UAV monitoring system using deep learning approaches. IEEE Trans Reliab. 2021;71(2):973–83.10.1109/TR.2021.3119068Search in Google Scholar

[5] Li X, Savkin AV. Networked unmanned aerial vehicles for surveillance and monitoring: a survey. Future Internet. 2021;13(7):174.10.3390/fi13070174Search in Google Scholar

[6] Valente J, António J, Mora C, Jardim S. Developments in image processing using deep learning and reinforcement learning. J Imaging. 2023;9(10):207.10.3390/jimaging9100207Search in Google Scholar PubMed PubMed Central

[7] Tan CH, Ng M, Shaiful DS, Win SK, Ang WJ, Yeung SK, et al. A smart unmanned aerial vehicle (UAV) based imaging system for inspection of deep hazardous tunnels. Water Pract Technol. 2018;13(4):991–1000.10.2166/wpt.2018.105Search in Google Scholar

[8] Shakhatreh H, Sawalmeh AH, Al-Fuqaha A, Dou Z, Almaita E, Khalil I, et al. Unmanned aerial vehicles (UAVs): a survey on civil applications and key research challenges. IEEE Access. 2019;7:48572–634.10.1109/ACCESS.2019.2909530Search in Google Scholar

[9] McEnroe P, Wang S, Liyanage M. A survey on the convergence of edge computing and AI for UAVs: opportunities and challenges. IEEE Internet Things J. 2022;9(17):15435–59.10.1109/JIOT.2022.3176400Search in Google Scholar

[10] Goodrich MA, Morse BS, Gerhardt D, Cooper JL, Quigley M, Adams JA, et al. Supporting wilderness search and rescue using a camera‐equipped mini UAV. J Field Robot. 2008;25(1–2):89–110.10.1002/rob.20226Search in Google Scholar

[11] Redmon J. You only look once: unified, real-time object detection. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2016.10.1109/CVPR.2016.91Search in Google Scholar

[12] Anagnostopoulos CN, Anagnostopoulos IE, Psoroulas ID, Loumos V, Kayafas E. License plate recognition from still images and video sequences: a survey. IEEE Trans Intell Transp Syst. 2008 May;9(3):377–91.10.1109/TITS.2008.922938Search in Google Scholar

[13] Niu H, Gonzalez-Prelcic N, Heath RW. A UAV-based traffic monitoring system-invited paper. In 2018 IEEE 87th Vehicular Technology Conference (VTC Spring). IEEE; 2018 Jun. p. 1–5.10.1109/VTCSpring.2018.8417546Search in Google Scholar

[14] Ranjan A, Misra P, Vasan A, Krishnakumar S, Sivasubramaniam A. City scale monitoring of on-street parking violations with streethawk. In Proceedings of the 6th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation. 2019 Nov. p. 31–40.10.1145/3360322.3360841Search in Google Scholar

[15] Hashesh AO, Hashima S, Zaki RM, Fouda MM, Hatano K, Eldien AS. AI-enabled UAV communications: challenges and future directions. IEEE Access. 2022 Aug;10:92048–66.10.1109/ACCESS.2022.3202956Search in Google Scholar

[16] Biørn-Hansen A, Rieger C, Grønli TM, Majchrzak TA, Ghinea G. An empirical investigation of performance overhead in cross-platform mobile development frameworks. Empir Softw Eng. 2020 Jul;25:2997–3040.10.1007/s10664-020-09827-6Search in Google Scholar

[17] Chaubey C, Sharma A. The integrated development environment (IDE) for application development: Android studio and its tools. In AIP Conference Proceedings. Vol. 2427, No. 1. AIP Publishing; 2023.10.1063/5.0116494Search in Google Scholar

[18] Olivier P, Chiba D, Lankes S, Min C, Ravindran B. A binary-compatible unikernel. In Proceedings of the 15th ACM SIGPLAN/SIGOPS International Conference on Virtual Execution Environments. 2019 Apr. p. 59–73.10.1145/3313808.3313817Search in Google Scholar

[19] Kamarudin KR, Wei YJ. Production of drone orthomosaic map of UTHM Wetland Conservation Research Station using UAV photogrammetry. In IOP Conference Series: Earth and Environmental Science. Vol. 1064, No. 1. IOP Publishing; 2022. p. 012012.10.1088/1755-1315/1064/1/012012Search in Google Scholar

[20] Terry BB, Baptist JR, Rattz J, Wagner O, Yetkin O, Gowda S. An end-to-end pipeline for acquiring, processing, and importing UAS data for use in the Open Data Cube (ODC). In 2021 IEEE International Geoscience and Remote Sensing Symposium IGARSS. IEEE; 2021 Jul. p. 8197–200.10.1109/IGARSS47720.2021.9553149Search in Google Scholar

[21] Fraser RH, Leblanc SG, Prevost C, van der Sluijs J. Towards precise drone-based measurement of elevation change in permafrost terrain experiencing thaw and thermokarst. Drone Syst Appl. 2022;10(1):406–26.10.1139/dsa-2022-0036Search in Google Scholar

[22] Bishop J, Horspool N. Cross-platform development: software that lasts. Computer. 2006;39(10):26–35.10.1109/MC.2006.337Search in Google Scholar

[23] Pang B, Nijkamp E, Wu YN. Deep learning with tensorflow: a review. J Educ Behav Stat. 2020;45(2):227–48.10.3102/1076998619872761Search in Google Scholar

[24] Smilkov D, Thorat N, Assogba Y, Nicholson C, Kreeger N, Yu P, et al. Machine learning for the web and beyond. Proceedings of Machine Learning and Systems. Vol. 1, 2019. p. 309–21.Search in Google Scholar

[25] Shastry S, Gunasheela G, Dutt T, Vinay DS, Rupanagudi SR. “i” – A novel algorithm for optical character recognition (OCR). In 2013 International Mutli-Conference on Automation, Computing, Communication, Control and Compressed Sensing (iMac4s). IEEE; 2013 Mar. p. 389–93.10.1109/iMac4s.2013.6526442Search in Google Scholar

[26] Win NN, Kyaw K, Win T, Aung P. Image noise reduction using linear and nonlinear filtering techniques. Int J Sci Res Publ. 2019 Aug;9(8):816–21.10.29322/IJSRP.9.08.2019.p92113Search in Google Scholar

[27] Liu YH. Feature extraction and image recognition with convolutional neural networks. In Journal of Physics: Conference Series. Vol. 1087. IOP Publishing; 2018 Sep. p. 062032.10.1088/1742-6596/1087/6/062032Search in Google Scholar

[28] Al-Shemarry MS, Li Y, Abdulla S. Detecting distorted vehicle licence plates using novel preprocessing methods with hybrid feature descriptors. IEEE Intell Transp Syst Mag. 2023;15(2):6–25.10.1109/MITS.2022.3210226Search in Google Scholar

[29] Yuheng S, Hao Y. Image segmentation algorithms overview. arXiv preprint arXiv:1707.02051; 2017.Search in Google Scholar

[30] Petrlík M, Báča T, Heřt D, Vrba M, Krajník T, Saska M. A robust UAV system for operations in a constrained environment. IEEE Robot Autom Lett. 2020 Feb;5(2):2169–76.10.1109/LRA.2020.2970980Search in Google Scholar

[31] Alcarria R, Bordel B, Manso MA, Iturrioz T, Pérez M. Analyzing UAV-based remote sensing and WSN support for data fusion. In Proceedings of the International Conference on Information Technology & Systems (ICITS 2018). Springer International Publishing; 2018. p. 756–66.10.1007/978-3-319-73450-7_71Search in Google Scholar

Received: 2024-08-09

Revised: 2024-10-12

Accepted: 2024-12-17

Published Online: 2025-02-22

This work is licensed under the Creative Commons Attribution 4.0 International License.

Articles in the same Issue

https://doi.org/10.1515/eng-2024-0101

Keywords for this article

AI; image processing; UAVs; OCR; systems integration

Creative Commons

BY 4.0