Spectrum

What it does

Spectrum enables real-time mangrove monitoring by converting fast RGB drone images into rich spectral data. Unlike HSI/MSI systems that take up to 30s per image, Spectrum processes RGB in real-time and visualizes insights instantly via a smart dashboard.

Your inspiration

Spectrum was inspired by a challenge from the UAE Space Agency. Traditional Hyperspectral (HSI)/Multispectral (MSI) systems take 20–30 seconds to capture a single image and produce large files that are hard to transfer, limiting real-time monitoring. This prompted us to find a faster solution that maintains spectral richness. We designed Spectrum to reconstruct HSI/MSI-like data from RGB images captured by drones and sent over 4G/5G for instant AI analysis and dashboard visualization—bridging the gap between speed, accuracy, and scalability.

How it works

Spectrum works by using artificial intelligence to convert regular RGB images into data-rich representations similar to those captured by expensive hyperspectral and multispectral cameras. We developed two AI models: one reconstructs 224-band hyperspectral data using a U-Net architecture trained on the open-source HySpecNet-11k dataset, and the other reconstructs 4-band multispectral data using a lightweight CNN trained on the UAE Space Agency’s GiQ dataset. These images are then analyzed for vegetation features and classified through a hybrid model. The results are visualized on a custom-built, cross-platform dashboard that presents real-time insights. On the hardware side, we built a prototype using a drone equipped with a standard RGB camera and an embedded controller, simulating the behavior of an industrial drone system capable of 4G/5G image transmission for instant processing.

Design process

After being inspired by the UAE Space Agency's challenge, we began by researching existing technologies to identify the exact market gap. We found that while hyperspectral imaging offers detailed data, it’s too slow and expensive for real-time environmental monitoring. With this in mind, we brainstormed ideas and sketched the full system concept on paper, visualizing how the interface would look and how each component would interact. We then moved into the prototyping phase. On the software side, we trained and fine-tuned multiple AI models for RGB-to-HSI and RGB-to-MSI reconstruction, optimizing for speed and accuracy. These models were trained on real hyperspectral datasets, including HySpecNet and GiQ, and were capable of producing high-fidelity outputs. For the hardware prototype, we built a drone system equipped with a standard RGB camera and embedded processor, simulating the architecture of industrial drones. This setup enabled high-speed RGB image capture and transfer over 4G/5G. The captured images were then linked to a custom cross-platform dashboard that we designed to provide real-time classification and easy visualization of mangrove health and types.

How it is different

Spectrum stands out by delivering the benefits of hyperspectral and multispectral imaging (rich environmental insights) using only RGB images. Unlike traditional systems that require costly, bulky cameras and take 20–30 seconds to capture a single image, Spectrum reconstructs high-fidelity spectral data in seconds using lightweight AI models. This enables real-time processing, classification, and visualization without specialized hardware. The system is compact enough to run on drones with standard cameras and embedded controllers, making it cost-effective, fast to deploy, and scalable. Additionally, Spectrum includes a custom dashboard that allows users to instantly view classification results and vegetation insights. Most existing solutions focus solely on raw data capture; Spectrum bridges the gap between capture, processing, and interpretation—delivering actionable insights from end to end.

Future plans

We are currently incubated at the Ajman University Innovation Center, where we’re transforming Spectrum from a prototype into a market-ready product. Our next steps include registering the intellectual property (IP) of the technology, refining the software and hardware systems for broader deployment, and expanding its use beyond mangroves into agriculture and climate monitoring. We aim to collaborate with environmental agencies and space programs to scale Spectrum into a cost-effective, real-time monitoring solution across diverse ecosystems.

Awards

- Top 10 winner in the UAE Hackathon 2025 - Incubated by the Ajman University Innovation Center (AUIC)