Scanworld aimed to provide farmers with accurate crop information using satellite imagery, particularly hyperspectral images. These images, rich in data, are ideal for assessing vegetation health, soil types, and crop maturity. However, hyperspectral images are costly to acquire, and they suffer from low spatial resolution, making them less effective for detailed analysis. Meanwhile, lower-cost multispectral images, which have better spatial resolution, lack the necessary data content for accurate crop predictions. Scanworld needed a solution to combine the advantages of both types of imagery to create more detailed, cost-effective, high-resolution hyperspectral images.
Working closely with Scanworld, we developed AI models capable of merging low-resolution hyperspectral images with high-resolution multispectral images. This collaboration focused on creating a model that could infer a high-resolution hyperspectral image from a multispectral one, enabling precise crop analysis and predictions without the hefty cost of traditional hyperspectral imagery.
More information about the project can be found in this article: “Agriculture: can Hyperspectral Imagery change the game?”
Sagacify's AI solution involved the creation of algorithms that integrate multispectral and hyperspectral data, allowing for a seamless transformation of low-resolution images into high-quality, high-resolution hyperspectral images. This model, unique in its learning-based approach, does not require prior knowledge about the target image and can be generalized across various scenes. The model utilizes the spectral information from multispectral images to accurately extrapolate 200+ spectral bands, delivering a high-resolution hyperspectral image with improved spatial accuracy and spectral resolution.
Feeling lost ?
Let's discuss over a coffee; our AI experts have you covered!
