David G. Goodenough

On the Slope-Aspect Correction of Multispectral Scanner Data

SUMMARYThe effects of topography on the radiometric properties of multispectral scanner (MSS) data are examined in the context of the remote sensing of forests in mountainous regions. The two test areas considered for this study are located in the coastal mountains of British Columbia, one at the Anderson River near Boston Bar and the other at Gun Lake near Bralorne. The predominant forest type at the former site is Douglas fir, whereas forest types at the latter site are primarily lodgepole pine and ponderosa pine. Both regions have rugged topography, with elevations ranging from 330 to 1100 metres above sea level at Anderson River and from 750 to 1300 metres above sea level at Gun Lake.Lambertian and non-Lambertian illumination corrections are formulated, taking into account atmospheric effects as well as topographic variations. Terrain slope and aspect values are determined from a digital elevation model and atmospheric parameters are obtained from a model atmosphere computation for the solar angles an...

Local maximum filtering for the extraction of tree locations and basal area from high spatial resolution imagery.

In this study we investigate the use of local maximum (LM) filtering to locate trees on high spatial resolution (1-m) imagery. Results are considered in terms of commission error (falsely indicated trees) and omission error (missed trees). Tree isolation accuracy is also considered as a function of tree crown size. The success of LM filtering in locating trees depends on the size and distribution of trees in relation to the image spatial resolution. A static-sized 3×3 pixel LM filter provides an indication of the maximum number of trees that may be found in the imagery, yet high errors of commission reduce the integrity of the results. Variable window-size techniques may be applied to reduce both the errors of commission and omission, especially for larger trees. The distribution of the error by tree size is important since the large trees account for a greater proportion of the stand basal area than the smaller trees. An investigation of the success of tree identification by tree crown radius demonstrates the relationship between image spatial resolution and LM filtering success. At an image spatial resolution of 1 m, a tree crown radius of 1.5 m appears to be the minimum size for reliable identification of tree locations using LM filtering.

Processing Hyperion and ALI for forest classification

Hyperion (a hyperspectral sensor) and the Advanced Land Imager (ALI) (a multispectral sensor) are carried on the National Aeronautics and Space Administrations Earth Observing 1 (EO-1) satellite. The Evaluation and Validation of EO-1 for Sustainable Development (EVEOSD) is our project supporting the EO-1 mission. With 10% of the worlds forests and the second largest country by area in the world, Canada has a natural requirement for effective monitoring of its forests. Eight test sites have been selected for EVEOSD, with seven in Canada and one in the United States. Extensive fieldwork has been conducted at four of these sites. A comparison is made of forest classification results from Hyperion, ALI, and the Enhanced Thematic Mapper Plus (ETM+) of Landsat-7 for the Greater Victoria Watershed. The data have been radiometrically corrected and orthorectified. Feature selection and statistical transforms are used to reduce the Hyperion feature space from 198 channels to 11 features. Classes chosen for discrimination included Douglas-fir, hemlock, western redcedar, lodgepole pine, and red alder. Overall classification accuracies obtained for each sensor were Hyperion 90.0%, ALI 84.8%, and ETM+ 75.0%. Hyperspectral remote sensing provides significant advantages and greater accuracies over ETM+ for forest discrimination. The EO-1 sensors, Hyperion and ALI, provide data with excellent discrimination for Pacific Northwest forests in comparison to Landsat-7 ETM+.

Spatial thresholds, image-objects, and upscaling: A multiscale evaluation

Abstract When examining a remotely sensed signal through various scale changes, what is the most appropriate upsealing technique to represent this signal at different scales? And how can this be validated? Solutions to these questions were approached by examining how the 660 nm signal of six forest stands vary through four different scales of same-.sensor imagery, four traditional resampling techniques, and a new object-specific resampling technique. Analysis of the original and modeled datasets suggests that appropriately upscaled imagery represents a more accurate scene-model than an image obtained at the upscaled resolution. Results further indicate the need for a multiscale approach to feature extraction and upscaling, as no single spatial resolution of imagery appears optimal for detecting or Upscaling the varying sized, shaped, and spatially distributed objects within a scene. By employing the human eye as a model, we describe a novel object-specific approach for addressing this challenge. Upscaling evaluation is based on visual interpretation, an understanding of the applied resampling theories, and the root mean square error results of 6000 samples collected from a 10 in CASI scene, and from 1.5 m, 3 in, and 5 m same site CASI images upscaled to 10 m. Potential application of this object-specific approach in hierarchical ecosystem modeling is also briefly described.

Compact Decomposition Theory

In this letter, we develop several new aspects of target decomposition theory for use with compact-mode polarimetric radar data. We first make a general link between fully polarimetric systems and compact modes before developing two important types of decomposition, namely, entropy/alpha and model-based surface/dihedral/volume techniques. We show that, under certain assumptions, compact data can be used to estimate the rotation invariant alpha angle of quadpol systems, which can then be used for polarimetric classification and physical parameter estimation. We apply the new methods to the problem of historical forest fire scar detection, using data at L- and C-bands to demonstrate the preservation of signatures in transition from quad to compact modes.

An Expert System for Remote Sensing

The Canada Centre for Remote Sensing has developed two hierarchical expert systems, the Analyst Advisor and the Map Image Congruency Evaluation (MICE) advisor. These expert systems are built upon our Remote-Sensing Shell (RESHELL) written in Logicwares MPROLOG. A shell is a programming environment that specifically caters to expert system development. Knowledge is represented in the production rules and frames database. Numerical processing takes place using the extensive FORTRAN code of the Landsat Digital Image Analysis System (LDIAS). The LDIAS includes several DEC VAX computers, image displays, specialized processors, and DEC Al VAXstations. The paper describes the architecture of the expert system to compare maps and images (MICE) and the expert system to advise on the extraction of resource information from remotely sensed data, the Analyst Advisor. Details are given concerning the structure of RESHELL and our methods of interfacing symbolic reasoning in PROLOG on the Al VAX stations with numeric processing in FORTRAN on several different computers. The first prototype of the Analyst Advisor will be released for internal use at CCRS in March 1987.

Nonlinear feature extraction of hyperspectral data based on locally linear embedding (LLE)

Feature extraction is an indispensable preprocessing step for information extraction from hyperspectral remote sensing data. In this paper, we introduce a nonlinear feature extraction algorithm, called Locally Linear Embedding (LLE), and customize it for hyperspectral remote sensing applications. Unlike the linear feature extraction algorithms based on eigenvectors of data covariance matrix, LLE preserves local topology of hyperspectral data in the reduced space. This preservation is important to maintain the nonlinear properties of the input data that benefits further information extraction. To investigate its effectiveness for hyperspectral remote sensing applications, LLE was examined in terms of spatial information preservation and pure pixel identification. The preliminary result of this study demonstrated that it compared favorably with PCA on spatial information preservation. In addition, it exceeded PCA on pure pixel identification through scatter plots. Index Terms – feature extraction, hyperspectral, dimensionality reduction, Principal Component Analysis, Locally Linear Embedding, information content.

Investigation of Nonlinearity in Hyperspectral Imagery Using Surrogate Data Methods

Although hyperspectral remotely sensed data are believed to be nonlinear, they are often modeled and processed by algorithms assuming that the data are realizations of some linear stochastic processes. This is likely due to the reason that either the nonlinearity of the data may not be strong enough, and the algorithms based on linear data assumption may still do the job, or the effective algorithms that are capable of dealing with nonlinear data are not widely available. The simplification on data characteristics, however, may compromise the effectiveness and accuracy of information extraction from hyperspectral imagery. In this paper, we are investigating the existence of non- linearity in hyperspectral data represented by a 4-m Airborne Visible/Infrared Imaging Spectrometer image acquired over an area of coastal forests on Vancouver Island. The method employed for the investigation is based on the statistical test using surrogate data, an approach often used in nonlinear time series analysis. In addition to the high-order autocorrelation, spectral angle is utilized as the discriminating statistic to evaluate the differences between the hyperspectral data and their surrogates. To facilitate the statistical test, simulated data sets are created under linear stochastic constraints. Both simulated and real hyperspectral data are rearranged into a set of spectral series where the spectral and spatial adjacency of the original data is maintained as much as possible. This paper reveals that the differences are statistically significant between the values of discriminating statistics derived from the hyperspectral data and their surrogates. This indicates that the selected hyperspectral data are nonlinear in the spectral domain. Algorithms that are capable of explicitly addressing the nonlinearity are needed for processing hyperspectral remotely sensed data.

Error reduction methods for local maximum filtering of high spatial resolution imagery for locating trees

Tree crown recognition using high spatial resolution remotely sensed imagery provides useful information relating the number and distribution of trees in a landscape. A common technique used to identify tree locations uses a local maximum (LM) filter with a static-sized moving window. LM techniques operate on the assumption that high local radiance values represent the centroid of a tree crown. Although success has been found using LM techniques, various authors have noted the introduction of error through the inclusion of falsely identified trees. Missing trees, or omission error, are primarily the result of too coarse an image spatial resolution in relation to the size of the trees present. Falsely indicated trees (commission error) may be removed through image processing post-LM filtering. In this paper, using 1-m spatial resolution multi-detector electro-optical imaging sensor (MEIS-II) imagery of a study location on Vancouver Island, British Columbia, we present a variety of techniques for addressing commission error when applying an LM technique. Methods exploiting spatial and spectral information are applied. As a benchmark, LM generated within a 3 × 3 window with no commission error reduction resulted in a 67% overall accuracy, with a 22% commission error. The results of the commission error reduction must be considered against resultant overall accuracy. Using variable window sizes, as suggested by image spatial structure, for the generation of LM provided for the maintenance of similar overall accuracy (62%) with a decrease in commission error (to 11%).

The CCRS SAR/MSS Anderson River Data Set

Technical Committee no. 7 of the International Association of Pattern Recognition is seeking test data sets that would further research into pattern recognition for remote sensing. Such data sets are usually expensive to acquire and are rarely made available. The Canada Centre for Remote Sensing (CCRS) has chosen to make the SAR/MSS Data Set for Anderson River available. This paper describes the contents and structure of the data set. Several major studies were conducted using these data by the authors and their colleagues. This paper will also summarize the results of these investigations conducted over four years. Studies included classification accuracies with and without terrain slope and aspect corrections, optimum sensor and feature selection, texture features, and multisensor data integration. Finally, the authors describe the procedure whereby other scientists can gain access to the data set.