Geographic Resource Solutions

Download

This download represents a presentation made at the May 2011 American Society of Photogrammetry and Remote Sensing (ASPRS)  Annual Conference in Milwaukie, WI.  It covers the basic design and implementation of the Accuracy Assessment undertaken at Lassen Volcanic National Park during 2008-09.  This assessment was designed to test two different maps (on based on photo-interpretation and the other on image classification) at the same time.  Each map was assessed using a stratified random sample design.  A data set of randomly of 25,000 randomly generated X,Y locations representing candidate accuracy assessment sites was generated.  These sites were spatially related to the map polygons they fell within.  List sampling was implemented by NVCS strata to select individual sample sites within each designated stratum.  Some points sampled only one of the maps while some points sampled both maps.  No points were moved away from boundaries, but the orientation of sample areas was altered in an unbiased systematic fashion so that all sample areas fell within the appropriate sample polygons.  Approximately 900 sites were visited and over 1000 land cover assessments were made - some sites had dual (different) samples to satisfy the need for independence between the sampling of the two maps.  Examples of list sampling queries and database updates are provided that demonstrate how to apply these techniques. Contact Ken Stumpf at GRS for additional information.

Download

Land Cover Mapping in Alaska Based on Satellite Image Classification –Lessons Learned.  Presented at the 2011 Alaska Survey and Mapping Conference, Anchorage, Alaska

Download

Brown, G. K., Fox, L.; "Digital Classification of Landsat Thematic Imagery for Recognition of Wildlife Habitat Characteristics"

Download

Hill, T. 1996. ''Forest Biometrics from Space''

In 1996 Geographic Resource Solutions (GRS) completed mapping the Applegate River watershed in southern Oregon for existing vegetation. This project was a cooperative effort between the USDA Forest Service, and Bureau of Land Management. GRS used Landsat TM satellite imagery, Digital Elevation Models, measured field data, GIS, and GPS. The final database estimated polygon attributes using continuous variables including canopy closure, average tree size, species composition, trees per acre, and variance for tree size and canopy closure. This paper describes the various methodologies used in the project, that include: field data collection, image processing techniques for removing the effects of topography, hybrid supervised and unsupervised image classification techniques, ecological rule-based pixel aggregation, and quantitative accuracy assessment.

Download

Stumpf, K. A. "Forest Inventory Stratification based on the Classification of Landsat Thematic Mapper Imagery"

Download

Stumpf, K.A. ''Image Processing Methodology.'' A slide presentation detailing the work involved in an image processing project from data acquisition to image classification to accuracy assessment.

Download

Stumpf, K.A. 1993. ''From Pixels to Polygons: the Rule-based Aggregation of Satellite Image Classification Data Using Ecological Principles''

Raster pixel data developed using satellite image classification techniques are frequently difficult to convert to a polygon (vector) format due to the extreme heterogeneity of the pixel classification data. Many groups of pixels are too small to map as polygons without yielding an unusable database. The small areas that are less than a user defined minimum size mapping unit must be aggregated with neighboring groups (stands) prior to developing a usable vector database. Conventional spatial operators based on either grid or polygon analysis (neighborhood and/or sliver filters) often cause degradation of stand boundaries and descriptive attributes, and decrease the reliability of the final map. A rule-based pixel filtering methodology that is based on user-defined concepts of stand similarity is presented in this paper. This technique considers stand characteristics such as the major vegetation type, species composition, density of canopy closure, average tree size, and canopy structure during the evaluation of stand similarity. Aggregation rules representing ecological relationships, minimum size constraints, and the relative importance of the different vegetation characteristics are also used to guide the aggregation process. The rules are flexible and may be defined relative to project objectives and the desired use of the resulting database.

Presented at the GIS'93 Conference, Vancouver, BC

Download

Stumpf, K.A. ''An Image Classification Sampling Methodology based on the Integration of IP/GIS Capabilities.'' A slide presentation that represents GRS's approach to developing image classification training site locations. Images are stratified.  Large homogeneous areas of the same class are identified.  A candidate site database is generated.  Frequency distributions that identify large frequently occurring classes can be distinguished from rare infrequently occurring classes.  Field sampling plans can be developed to represent the diversity of classes across the landscape without oversampling frequently occurring classes.

Initially presented at the 2002 ASPRS Conference in Washington, D.C.; also presented at the 2004 ASPRS Conference, Denver, CO

Download

Stumpf, K.A. ''The Aggregation of Pixel Data into Mapped Area Features.'' A slide presentation that contrasts inappropriate approaches of pixel aggregation, such as filtering and smoothing with more accurate and effective techniques like ecological rule-based pixel aggregation

Download

Hill, T. 1993. ''Taking the " " Out of "Ground Truth": Objective Accuracy Assessment'' Presented at the 12th Pecora Conference, Sioux Falls, SD