Meet Hyperkon: Hyperspectral-ready Convolutional Neural Network
Existing computer vision techniques primarily designed for RGB images need to be adapted to this high-dimensional hyperspectral modality. We developed HyperKon, a hyperspectral-ready Convolutional Neural Network (CNN) backbone designed for hyperspectral data and trained on a substantial custom dataset comprised of HyperSpectral Image (HSI).
HyperKon can learn meaningful representations from hyperspectral data without the need for extensive labelled data. HyperKon has the potential to improve performance across a wide range of hyperspectral image analysis tasks, including hyperspec- tral super-resolution and classification.
Committed to advancing the capabilities of hyperspectral image analysis, Sixteen Sands continues to refine and extend the functionality of HyperKon. Our ongoing research focuses on optimizing HyperKon for diverse applications, such as fine-grained hyperspectral super-resolution and robust classification tasks.
Capture and Process Data
Hyperspectral sensors are devices that capture and process data throughout the electromagnetic spectrum using a sequence of streamlined and adjacent bands. This capability allows them to capture the distinctive spectral signatures of objects and materials, revealing their composition and identity.
Precision Agriculture
Hyperspectral Sensor Imaging finds applications in various fields, including astronomy, agriculture, geology, biology, and surveillance. In agriculture, which is our core focus hyperspectral data can support farmers in optimizing resource usage, detecting diseases and pests, monitoring crop growth and yield, and assessing soil quality and fertility.
Building with Unmanned Aerial Vehicles
Sixteen Sands is creating techniques that can harness the potential of hyperspectral data acquired by Unmanned Aerial Vehicles (UAVs). Our research investigates Computer Vision techniques that allow enhancing the spectral information in affordable UAV multispectral data using coarse spatial-resolution satellite HSI.
As we delve into the vast possibilities of building with Unmanned Aerial Vehicles, our commitment to pushing the boundaries of technology remains unwavering. Through collaborative partnerships and interdisciplinary approaches, Sixteen Sands aims to make significant contributions to the growing field of UAV-based remote sensing, opening up new avenues for impactful research and development.
Continuously evolving in this rapidly changing landscape, Sixteen Sands remains dedicated to staying at the forefront of technological advancements. By embracing cutting-edge methodologies and fostering a culture of exploration, we aim to redefine the possibilities within UAV-based hyperspectral remote sensing.
High Quality Data
Our research will leverage the existence of high-quality Hyperspectral sensors to investigate their use for extracting high-quality precision agriculture data using Deep Learning techniques. Several studies have established the feasibility of integrating multispectral imagery with hyperspectral data using deep learning techniques for enhanced downstream tasks. Keeping in line with these advancements, we aim to delve into the application of Vision-Based AI techniques to HSI, with the end goal of enhancing Sustainable Agricultural Management practices.
Digital Agriculture
Specifically for agriculture, hyperspectral sensors can measure the reflected radiation from plants in narrow and continuous spectral bands, which helps to obtain detailed information about their biochemical and physical properties
