Скачать или смотреть QGIS + Streamlit + Deepseek R1 Reasoning AI for NDVI and Topographic Data Analysis

In this video, Jesse discusses a project focused on vegetation assessments using drone-based NDVI imagery while accounting for topography. They explain their process of classifying landscapes into topographic positions—hilltop, mid slope, and depression—using high-resolution ortho images and digital surface models. The presenter utilizes a Gaussian weighting function to create a TPI raster, which is then converted into polygons for statistical analysis. They demonstrate a workflow built in QGIS model designer to streamline this process.

The analysis involves running zonal statistics on the NDVI data across different topographic zones and experimenting with various bandwidths. The presenter also showcases a web app developed using Replit that allows for efficient data handling and visualization of results. They highlight trends observed in NDVI values across different zones and resolutions, noting that low areas generally exhibit higher values than hilltops.

Additionally, they explore integrating AI reasoning models to analyze data outputs more effectively, suggesting potential applications for future investigations.
There is great opportunity to move AI interfaces beyond the typical chatbot - I this example we see how to use python to do data analysis and feed that data along with a detailed system prompt into a reasoning model like Deepseek R1 to get a simple report output.

Overall, the project combines GIS techniques with AI tools to enhance vegetation assessment methodologies.