Materials detection technology gives you the tools to identify the location and extent of natural and man-made materials on Earth’s surface.
Using multispectral data to identify materials by their unique spectral “fingerprints,” you can accurately locate and identify materials that the human eye cannot see. With GBDX, DigitalGlobe’s cloud-based platform for geospatial big data, you can cost-effectively execute sophisticated detections, in multiple locations and over large geographic areas. Materials detection can solve many problems. But no matter what materials you need to identify or visualize, you’ll need to output the data in a format designed for your analytics.
Some possible uses for materials detection include:
- Determine roofs’ potential for solar panel installation.
- Identify individual minerals, combinations of minerals, or patterns of minerals that might indicate subsurface ore deposits in order to recommend the highest probability locations to drill.
- Predict where acid mine drainage is likely to happen or monitor an abandoned mine or mill site for changes in minerals or materials that might indicate a pending spill or flood.
- Assess insurance risk by surveying roof materials for their flammability.
DigitalGlobe’s multispectral satellite imagery measures the reflective properties of all the materials on Earth’s surface, in far more detail than the human eye can see. Some of our sensors take in eight spectral channels; others take in sixteen. By measuring the reflective properties of materials in different channels, we can compare what we observe in these images to what we know about spectral channels for different materials.
To evaluate materials, you first request the imagery from our 15-year archive of high-resolution satellite imagery and normalize it through a standard set of algorithms we provide that are designed to normalize the data for further processing. Through processes such as orhtorectification and atmospheric compensation, we take years of expertise and expose these algorithms as simple tools for non-experts to leverage in preparing data for complex analysis. We turn the measurements made by our sensors into a reflectance image so that you’re ready to run a range of spectral detection algorithms based on traditional methods like Spectral Angle Mapper (SAM) or Matched filtering, or new-age methods based on computer vision and deep learning. You can export your results as a table or a cartographic output to observe the location and distribution of your target materials.
Ordering your Data
Data can be requested for analysis through the use of a simple set of REST APIs. First, have a look in our catalog via the catalog API to determine if your data is already in the cloud and ready for analysis. If the data has not yet hit our cloud processing system, you can easily request imagery for a particular area of interest (AOI) by ordering an image from our ordering API. Once you’ve ordered an image, the data will arrive in the platform within minutes to hours to perform analysis. You can search the entire DigitalGlobe archive, and order all the data you need to analyze. Once it arrive in the GBDX platform, you’re ready to perform your material detection analysis.
Our platform development team have built a framework to make it simple and straightforward to launch a set of algorithmic tasks against the imagery you want to analyze. A set of pre-defined tasks are available through the platform, and these tasks can be coupled to run in succession or in parallel through the use of our Workflow API.
In order to achieve the highest accuracy from the imagery, the images first need to be sent through a set of algorithms that clean it up and normalize it.
- We request the data using the Catalog and Ordering APIs, then run our atmospheric compensation algorithm to account for effects on the image including haze, aerosols, moisture, sun illumination angle and viewing geometry. The output of this process is a reflectance image that is a normalized set of units. Now we are ready to begin analyzing the imagery to map materials within the scene.
- Next, we compare the corrected measurements to a library of known material reflectance signatures. These signatures are often developed through careful measurements of target materials using a spectrometer, under controlled environments, at places like the USGS and in National Laboratories. A simple process to compare laboratory signatures with our reflectance imagery is called spectral angle mapper, or SAM. With SAM, we compare the spectral angle between two multidimensional vectors: the spectral angle observed in-scene and the spectral signature of each target material. After SAM, the data can be exported to a cartographic output, creating an abundance map for a particular material.
- In some cases we’ll perform some statistical analysis over the resulting classification image to report the materials that are most abundant within a location. For example, a structure may be comprised primarily of pixels that match our signature for shake shingle, so we report that structure as having a shake shingle roof.
Whether you’re looking for an abundance map or a table output, the sharp, shiny data is now ready for your use. The results from any GBDX workflow are exported to a user-specific S3 bucket where the data can be viewed, downloaded or incorporated into other data sets for further analysis.
In the home insurance industry, we’ve seen producers use GBDX to get highly accurate, up-to-date information about roof materials. With our algorithms, the insurer can observe and differentiate the risk profile of a property based on the information. The table output assigns each address a material type, showing the insurance agent the roof material information for each individual property.
Using this output, an insurer can vastly decrease the amount of manually-entered information required to calculate a new quote. The fewer questions a potential new customer has to answer and the shorter the time from initial click to quote, the higher the conversion rate, for example.
For the mining industry, GBDX provides the ability to detect visible outcroppings of particular rock units that contain valuable minerals. To the human eye, many rocks in our imagery all look the same. Using the reflectance signatures observed by our sensors, we can obtain far more detailed information about these exposures. Using SAM, we recognize and map the location and distribution of particular visible exposures with high concentrations of the desired minerals. In this case, the output is often an abundance map showing the location and quantity of the rock unit. The possibilities for leveraging rich geospatial information are endless. With 15 years of high-resolution, multispectral data on the GBDX platform, analysis of the data is accessible and ready for further refinement. There’s likely an optimal algorithm or sets of algorithms for roof material detection, specific mineral detection, vegetation species analysis, or a host of other material analysis use cases. We’re excited to see what data scientists, imagery experts and algorithm developers can build with the data, tools and platform now available.