Tim J. Patterson

Radiometrically correct sharpening of multispectral images using a panchromatic image

With the advent of a panchromatic band on the enhanced thematic mapper (ETM) sensor, the interest in sharpening techniques for multispectral imagery has substantially increased. Previous work has been performed to increase the spatial resolution of multispectral imagery by combining it with a higher spatial resolution broad band panchromatic source. While these techniques resulted in a sharper multispectral image, they limited the extent to which the spectral information could be used. In this paper, we describe a technique that employs pseudoinverse estimation to generate a sharpened multispectral image. This technique provides the minimum mean squared error estimate of the sharpened spectral signatures given the information available from the panchromatic and multispectral sensors. The resulting radiometrically correct sharpened image is useful for various types of spectral analysis such as spectral classification, demixing, and detection. We then present the results of a quantitative evaluation of the processing technique in various operating modes as well as the results of a qualitative evaluation from analysts using MSI for mapping purposes.

Automated training of 3D morphology algorithm for object recognition

Grayscale morphology has demonstrated a great deal of success in automatic target recognition (ATR) applications with a variety of imagery sources including SAR, IR, visible, and multispectral. However, training the morphology algorithm requires significant experience and is labor intensive. This paper presents an innovative approach for using genetic algorithms (GA) and the classification and regression trees (CART) algorithm to automate morphology algorithm training and optimize detection performance. The GA is used to find the morphology operators by encoding them into binary vectors. The CART algorithm determines the optimum region filtering parameters in conjunction with the morphology operations. Robustness is achieved by regression pruning of the CART generated classification trees. The basic concepts in applying the GA to the design of grayscale morphology filters is described. Our results suggest that the detection performance of a GA designed morphology filter is comparable to that designed by human experts. The automated design method significantly shortens the design process.

Extension of spatial sharpening techniques for hyperspectral data

This paper describes our approach and presents measured results of the extension of multispectral sharpening techniques to hyperspectral imagery. Our approach produce high spatial resolution spectral imagery using a least squares estimator. The estimator is based on the underlying physics of the spectral imaging process. The intent of the process it to produce high spatial resolutions with the best possible spectral fidelity. The results on multiple test cases demonstrate sharpened imagery within 5% of the true high resolution hyperspectral values.

Quantitative comparison of multispectral image-sharpening algorithms

This paper presents a quantitative comparison of a multispectral sharpening algorithm, which was introduced previously, with other standard techniques. The sharpening algorithms combine a high spatial resolution panchromatic image with a lower resolution multispectral image. The combination is based on a pseudo inverse of the image formation equations. The paper begins with an introduction motivating the technique. Previous approaches to the sharpening problem are then outlined. This is followed by a description of two new approaches. The first is an improvement on the standard intensity saturation hue (ISH) transform and the second is a sharpener which was introduced in previous papers. After descriptions of the sharpeners, the more important results of a series of experiments to evaluate the sharpener performance are presented. A full series of tests in which low resolution multispectral data was synthesized from a high resolution scene, sharpened with various techniques and compared to the original high resolution imagery was conducted. The most significant results are presented in this paper. A second test was conducted using both high and low resolution images collected of the same area. Sharpened low resolution multispectral images were compared to actual high resolution imagery of the same area.

Design of optimal transformations for multispectral change detection using projection pursuit

Effective change detection techniques for the automated detection of changes of interest have been an elusive goal for many years. The problem has never been one of detecting changes but, rather one of finding the changes of most interest among all the spurious changes. Indeed, changes of interest for one application can be completely different than the changes of interest for another application. In this paper we present a brief overview of techniques to suppress changes between scenes due to different collection conditions. Techniques for sorting the detected multispectral changes according to the intended application and relating them to actual changes on the ground are presented. Our approach is to use multispectral data transformations designed by the use of a visualization tool called Projection Pursuit. This tool allows the user to design a projection of the data into a vector space specifically designed to accentuate the visibility of the changes of interest. Hence, for change detection interactive analysis, projection pursuit offers the important advantage of being able to find the optimum projection without requiring a priori information from the image analyst and requiring little human intervention. This algorithm is complemented by canonical and principal component transformations tailored for specific exploitation requirements. The approach allows design of custom change detection products for a wide variety of applications including: military, economic, and environmental. This capability reduces the burden of data manipulation decisions required of the analyst, while still providing the flexibility required for the demands of exploitation.

Multisensor image cueing (MUSIC)

There have been many years of research and development in the Automatic Target Recognition (ATR) community. This development has resulted in numerous algorithms to perform target detection automatically. The morphing of the ATR acronym to Aided Target Recognition provides a succinct commentary regarding the success of the automatic target recognition research. Now that the goal is aided recognition, many of the algorithms which were not able to provide autonomous recognition may now provide valuable assistance in cueing a human analyst where to look in the images under consideration. This paper describes the MUSIC system being developed for the US Air Force to provide multisensor image cueing. The tool works across multiple image phenomenologies and fuses the evidence across the set of available imagery. MUSIC is designed to work with a wide variety of sensors and platforms, and provide cueing to an image analyst in an information-rich environment. The paper concentrates on the current integration of algorithms into an extensible infrastructure to allow cueing in multiple image types.

Hierarchical spatial/spectral processing for road mapping

This paper describes a work in progress to develop an updated road mapping system. The system is designed to generate map products working directly from multispectral imagery. The updated system is uses a resolution hierarchy to match the size of the roads, measured in image pixels, to the optimal processing configurations. The original system was designed to work with low resolution Landsat TM imagery while the updated system is designed to be more versatile with the ability to generate products from the new systems now available such as Landsat VII, IKONOS, and Digital Globe. The majority of the map production is performed in an automated mode requiring no user interaction. A Java interface to support final editing of the automated results has been built to support application on multiple platforms. This paper describes the mapping algorithms, the special editing interface designed for road vector maps, and results of some processing experiments.

Automated map generation and update from high-resolution multispectral imagery

This paper describes the LOCATE TNG system, which generates map products directly from multispectral imagery in an automated fashion. The LOCATE TNG system uses spectral and spatial feature information to extract various types of man- made lines of communication (LOCs) from imagery and generate them in the form of digital vector maps. The generated maps may be compared against reference digital maps to automatically find new or changed LOCs. The original LOCATE (lines of communication apparent from thematic mapper evidence) system was designed and developed to use landsat thematic mapper imagery having a resolution of 30 m. LOCATE TNG (the next generation) has been redesigned to also have the capability to use high resolution multispectral imagery to be available from the next generation of commercial satellites. These satellites will provide multispectral and panchromatic imagery having resolutions down to 4 m and 1 m, respectively, thus dramatically improving the information available for exploitation. LOCATE TNG employs a hierarchical algorithmic approach to extracting layers of LOCs (primary roads, secondary roads, etc.) that may be used for GIS applications.

Broad-area search for targets in SAR imagery with context-adaptive algorithms

This paper describes an ATR system based on gray scale morphology which has proven very effective in performing broad area search for targets of interest. Gray scale morphology is used to extract several distinctive sets of features which combine intensity and spatial information. Results of direct comparisons with other algorithms are presented. In a series of tests which were scored independently the morphological approach has shown superior results. An automated training systems based on a combination of genetic algorithms and classification and regression trees is described. Further performance gains are expected by allowing context sensitive selection of parameter sets for the morphological processing. Context is acquired from the image using texture measures to identify the local clutter environment. The system is designed to be able to build new classifiers on the fly to match specific image to image variations.

Perceptual grouping to extract extended linear features in Thematic Mapper (TM) imagery

In this paper we present a new method for the extraction of information about lines of communciation (LOC) from Landsat Thematic Mapper (TM) imagery. While other techniques have demonstrated some success in the extraction of LOC information, they generally suffer from the inability to handle a wide variety of LOC shapes or they lack the capability to globally integrate information in order to determine the location of the LOCs. In this paper we present the development of an alternative to previously tried methods which promises to do a better job of both local and global extraction of LOC information.