SBIR-STTR Award

Observera proposes to integrate several innovative fusion concepts (resolution enhancement technology, high-resolution signature data, and spatial texture) into the Army's process of Land Cover / Land Use (LCLU) classification from LANDSAT. This integration will result in substantially improved classification over current techniques. Specific focus will be on automating as much of these techniques as possible in order to maximize their utility for quick response missions. In addition, we will also develop the collection strategies necessary for cost-effective collection of the high-resolution imagery needed to support these new concepts. Demonstrating these capabilities lays an effective foundation for the development of an integrated classification process in Phase II, which will consist of the development of a prototype LCLU system. The Phase II effort will use rapid prototyping methodologies in order to develop a series of incrementally more capable prototype systems. Commercialization of the integrated system is one possible avenue, either as an add-on to a commercial image processing package or bundled with a commercial package as a complete system. Additionally, several of the individual components have independent commercial viability. In the larger context, the concepts developed here also enhance our ability to provide customized feature extraction capabilities to government and commercial customers worldwide.

Observera's objective for Phase II is to build upon the capabilities developed in Phase I and increase the accuracy and quality of LCLU classification results obtainable from LANDSAT through the innovative application of three techniques (Statistical Prediction of Bands, Spatial Spectral Bootstrapping and Texture Based Classification). Secondary objectives include increasing the efficiency of the overall workflow, providing compatibility with existing Army systems, and ensuring universal applicability of the process. Our Phase II effort will be directed toward building and testing this prototype system, called the Application for Land-use, Land-class Extraction for Geospatial Region Operations (ALLEGRO). The ALLEGRO prototype will be a user-friendly, semi-automated tool that will be compatible with existing Army terrain systems and will support the requirements of operational warfighters. ALLEGRO will address the primary factors limiting classification performance: mixed pixels, similar spectra, data variance, and training set errors. ALLEGRO will consist of a software application that performs the following functions: -Ingests LRMSI and HRMSI -Assists the operator in selecting optimal patches of HRMSI -Increases the resolution of the MSI using time-coincident panchromatic data -Performs semi-automated land-cover classification -Generates raster class-maps in ERDAS Imagine-compatible format. Commercialization of the integrated system is a primary avenue, either as an add-on to a commercial image processing package (such as ERDAS Imagine) or bundled with a commercial package as a complete system. Additionally, several of the individual components have independent commercial viability that will augment Observera's existing commercial product lines.

Keywords:
multispectral, landsat, ikonos, feature extraction, image processing, classification, resolution enhancement, land use