Accur Vision: Mobile application for 3D tracking solutions

Although the original developers used React Native for Web, we ran into its limits — the software was too complex and the web interface contained too many exceptions. We couldn't simply reuse the code for a mobile app.

During development, we had to verify our calculations in the real world. We implemented satellite images from Google Maps but encountered inaccuracies in zone rendering. For example, a parking lot was displayed in a different place than on the map. We used tennis courts with standardized dimensions to verify whether the error was in the application or the data.

In May 2024, we launched improvements to the mobile application: intruder movement on the map is significantly smoother, battery consumption decreased by approximately 10-15%, 96% fewer data downloads are required, and we know with greater accuracy that the server connection is stable.

The biggest challenge was the limits of React Native for Web — the software was too complex to simply port to mobile. We also had to solve the mismatch between virtual and physical worlds when rendering security zones on satellite maps.

The Cookielab team is working on a new web solution that will work more efficiently with data transfer. The goal is a reliable, responsive application easily integrated into other solutions like control room dashboards. We've also started developing native iOS and Android applications with improved UI and UX.

Accur8vision has a strong background as part of Hexagon, and its prestigious clientele includes businesses from a wide range of industries, critical infrastructure organizations, and security forces. Their project offers a mobile application providing a simplified view of the guarded area, automatically sending notifications about intruders. The company already had a web interface in React Native for Web, and they turned to us for mobile development.

Adding features like push notifications was relatively easy. The challenge was with maps. Accur8vision uses LiDAR to model intruder movement in 3D. For mobile, they generate a bird's-eye view where we draw intruders as points.

On the website, this view was written in Canvas — standard HTML technology that React Native doesn't support. We tried SVG, but the app crashed with large maps because SVG-to-bitmap conversion consumed too much memory. We considered CSS, but ultimately modified classic Canvas. We ran into problems finding polygon centers using standard calculations — we used an algorithm from Mapbox designed for such cases.

To render intruder movement in real-time, we call the server constantly — about 6 endpoints every two seconds. Communication wasn't prepared for this volume, and we brought down the entire server with one instance of the mobile app. We finally solved everything but plan to use gRPC for communication in the future.

We're working on a new web solution with more efficient data transfer. The goal is a reliable application that's responsive on different displays and easily integrated into dashboards. We plan to use gRPC for communication to handle real-time data volume more efficiently.