vPatrol from Tahaluf
— Written 2023 —
vPatrol is an AI-powered system that turns any fleet vehicle into an AI-powered, surveillance-enabled patrol fleet. It integrates cutting-edge AI vision technology with IoT architecture to provide real-time monitoring, analytics, and actionable insights.
MY ROLE
As the Product Owner for vPatrol, I led a multidisciplinary team through the entire product development lifecycle, from initial concept to the successful launch of the MVP. My role involved conducting comprehensive product discovery, where I collaborated with key stakeholders to define the core feature set that would deliver value for our customers.
Working closely with business analysts, AI experts, embedded systems engineers, data scientists, and backend developers, we designed and implemented a cutting-edge MVP tailored to the needs of large clients in the region. Our MVP successfully showcased vPatrol’s capabilities, providing a solid foundation for future product iterations and scaling.
PRODUCT
Equipped with advanced License Plate Recognition (LPR) technology, vPatrol allows vehicles to scan and recognize license plates of nearby vehicles while moving at speeds between 15-30 km/h. The system continuously sends plate information to a centralized platform, where it compares the data against a watchlist of wanted vehicles. If a match is found, an alert is immediately sent to the patrol officer, enabling rapid action.
In addition to vehicle recognition, vPatrol is equipped with AI-driven capabilities to detect crowds (more than five individuals) and identify fighting scenes or disturbances, enhancing situational awareness and providing a comprehensive approach to on-the-move security.
METRICS (MVP)
95% Accuracy in license plate recognition while moving at speeds between 15-30 km/h.
80% reduction in manual vehicle checks through automated License Plate Recognition (LPR), enabling officers to focus on more critical tasks.
Monitored and analyzed +100 vehicles, transmitting data to the central platform in real-time.
Crowd detection algorithms achieved a 90% accuracy rate in identifying gatherings of more than five people and potential disturbances such as fighting scenes.
KEY TAKEAWAYS
While leading the vPatrol product, I expanded my ability to work with cutting-edge technologies like AI and IoT, enhancing my product management capabilities without the need for hands-on coding. This reinforced my ability to lead multidisciplinary teams across technical domains.
Engaging regularly with field officers during product testing allowed us to refine features like crowd detection and vehicle flagging, ensuring vPatrol met the demands of real-world scenarios effectively.
Working in a fast-paced tech environment taught me the importance of managing resources efficiently. I learned how to work within the constraints of a budget while delivering an innovative, high-impact solution on time.
We faced many challenges that required constant iterations of the product design, frequent field tests, and continuous improvement of the AI algorithms, especially in managing environmental variability and ensuring the detection system worked reliably under various real-world conditions.
One of the major challenges we encountered was ensuring real-time detection while maintaining a high level of accuracy. Given that the system needed to recognize multiple license plates and crowds simultaneously while the vehicle was moving, this required optimizing the AI algorithms to process and relay data with minimal lag. Balancing detection speed with precision was crucial to avoid false positives or missing critical events, especially in high-traffic areas.
Daytime and nighttime conditions presented unique challenges for our detection algorithms. During the day, direct sunlight often causes glare on the license plates, making it difficult for the cameras to capture clear images. This led us to develop and fine-tune the system’s image correction algorithms to reduce the impact of bright plates or reflections from sunlight.
At night, the issue became more about low-light performance. Ensuring the cameras could detect vehicles and crowds accurately in poorly lit areas required using enhanced image recognition technologies. We had to continuously refine the system to function in urban areas with varying lighting conditions, including streetlights, headlights, and other light sources.
The system had to operate in diverse environments — from densely populated urban areas to quieter residential streets. Each environment presented unique challenges, such as moving through high-traffic zones, detecting larger crowds, or distinguishing between moving objects and static surroundings in a timely manner. It was important to fine-tune the object detection model so it would not be overwhelmed by irrelevant or excessive data from busy streets or crowded areas.
*This project is showcased for portfolio purposes only. All logos, trademarks, and brand names belong to their respective owners.