Table of Contents¶
- Hyperion data preparation
- Model training
- Model inference
- Results visualization
Generally, multispectral imagery is preferred for Landuse Landcover (LULC) classification, due to its high temporal resolution and high spatial coverage. With the advances in remote sensing technologies, hyperspectral images are now also a good option for LULC classification, due to their high spectral resolution. The main difference between multispectral and hyperspectral imagery, is the number of bands and how narrow those bands are. One of the advantages hyperspectral sensors have over multispectral sensors is the ability to differentiate within classes. For instance, due to the high spectral information content that creates a unique spectral curve for each class, hyperspectral sensors can distinguish by tree or crop species.
In this notebook, we will use hyperspectral data to train a deep learning model and will see if the model can extract subclasses of two LULC classes: developed areas and forests. Hyperion imagery is used in the current analysis to classify the types of forests and developed areas. The data can be downloaded from USGS earth explorer.
The Earth Observing-1 (EO-1) satellite was launched November 21, 2000 as a one-year technology demonstration/validation mission. After the initial technology mission was completed, NASA and the USGS agreed to the continuation of the EO-1 program as an Extended Mission. The EO-1 Extended Mission is chartered to collect and distribute Hyperion hyperspectral and Advanced Land Imager (ALI) multispectral products according to customer tasking requests.
Hyperion collects 220 unique spectral channels ranging from 0.357 to 2.576 micrometers with a 10-nm bandwidth. The instrument operates in a pushbroom fashion, with a spatial resolution of 30 meters for all bands. The standard scene width is 7.7 kilometers.
Login to the earth explorer using the USGS credentials. Then, select the Address/Place option in the Geocoding Method dropdown and write the name or address of the area of interest.