IMAGINE Subpixel Classifier™

Advanced Classification Simplified
IMAGINE Subpixel Classifier™ is a leading image classification tool for advanced spectral analysis and interpretation. It enables you to extract more information from your spectral imagery and dramatically improves your ability to find and quantify information at the whole and subpixel level.

IMAGINE Subpixel Classifier allows you to
· Detect a specific Material of Interest
· Automatically search for a specific Material of Interest in multiple scenes
· Generate purer spectral signatures and residuals
· Create more complete classifications
· Inventory the specific Material of Interest in an image

Overview
IMAGINE Subpixel Classifier is a state-of-the-art, supervised classifier that detects and reports whole and subpixel occurrences of a specific Material of Interest in multispectral imagery. IMAGINE Subpixel Classifier derives a signature for a specific material, groups all pixels in an image as either containing the material or not, and classifies all pixels that do contain the material into classes based on how much of the material they contain. It is a fully integrated, add-on application module to ERDAS IMAGINE.

Benefits
· Classify objects smaller than the spatial resolution of the sensor
· Discriminate specific materials in mixed pixels
· Detect materials that occupy from 100% to as little as 20% of a pixel
· Report the fraction of material present in each pixel classified
· Develop signatures portable from one scene to another
· Correct imagery for atmospheric effects
· Search broad-area images quickly to detect small or large features mixed with other materials

Features
· Powerful, low cost alternative to ground survey, field sampling, or high resolution imagery over wide areas or in remote locations
· Fully integrated with ERDAS IMAGINE
· Accepts airborne and satellite multispectral imagery such as Landsat Thematic Mapper, SPOT, IKONOS, and QuickBird multispectral imagery

Unique Solution
IMAGINE Subpixel Classifier is specifically designed to compensate for spectral mixing, atmospheric factors, and sun angle effects inherent to multispectral imagery. It applies an entirely different approach to background removal, environmental correction, and signature development. IMAGINE Subpixel Classifier takes you beyond the bounds of traditional classifiers and succeeds where alternative subpixel techniques fall short.

By addressing the mixed pixel problem, IMAGINE Subpixel Classifier can successfully detect a material even when it occupies only a fraction of an image pixel. As a bonus of the process, it tells you how much of the material is present in each pixel. You are aware not only of the material’s presence but also of the fraction of each image pixel it represents.

View a quantitative evaluation of IMAGINE Subpixel Classifier

View a White Paper on the IMAGINE Subpixel Classifier

Because IMAGINE Subpixel Classifier also addresses the spectral distortion of atmosphere and sun angle affects within an image, you can develop a material signature in one image and successfully apply it to others captured at different times and different places.

Technique
The IMAGINE Subpixel Classifier processing technique is a new approach to deriving spectral signatures and applying them to imagery. Traditional classification techniques develop signatures by combining the spectra of training set pixels. These signatures encompass the contributions of all the materials in the training set. In contrast, IMAGINE Subpixel Classifier’s Signature Derivation process extracts the component of the pixel spectra that is most common to the training set. The result is a signature that is representative of a specific material or set of materials.

Upon deriving a signature, conventional classifiers identify pixels in the scene that have the same spectral properties as the signature. IMAGINE Subpixel Classifier, however, locates pixels containing the signature as a fractional component of the overall spectrum. IMAGINE Subpixel Classifier performs a subpixel background estimation and removal process to produce a residual spectrum for each pixel in the scene. A detection occurs when the difference between the residual spectrum and the signature falls within a set of tolerances.

Suppression of subpixel background materials by IMAGINE Subpixel Classifier allows classifications of materials even when they occur in mixed pixels containing a wide range of background materials. IMAGINE Subpixel Classifier then reports the amount of the material associated with each pixel as both a histogram and graphic overlay.

Frequently Asked Questions regarding the IMAGINE Subpixel Classifier

IMAGINE Subpixel Classifier Outperforms other Classifiers

When to Use
IMAGINE Subpixel Classifier is designed as a complement to your existing suite of classification tools. It provides a higher level of discrimination for most applications. IMAGINE Subpixel Classifier is highly specific because its background removal function attempts to remove dissimilar materials from pixels with the intent to provide a discriminating and unique signature of the material of interest. The resulting signature is specialized to your MOI. Traditional classifiers attempt to include all materials in the pixels.