The accuracy of vegetation stand boundaries derived from image segmentation in a desert environment

Line intercept sampling was used to determine if boundaries between desert scrub vegetation stands corresponded with boundaries between regions in an image derived from a segmentation algorithm applied to Thematic Mapper (TM) data for the Anza-Borrego Desert State Park, California. An image segmentation algorithm developed by Woodcock and Harward (1992) was applied to images comprising TM bands 3, 4, and 5, from April 1987, principal components images based on April 1987 and June 1990 imagery, and each with texture added. The Global Positioning System (GPS) was used to determine coordinates of both physiographic (land-form) and vegetation boundaries in the field as they intersected line transects. These boundary locations were then registered to the segmented images. Image region boundaries that fell within E tolerances (spatial error bounds) of surveyed boundaries were considered accurate. Image region boundaries showed less than 10 percent omission error but about 50 percent commission error when compared with the true locations of vegetation and physiographic boundaries. The use of image principal components and texture in the segmenfations did not produce anticipated increases in the correspondence between field-mapped and image-region vegetation boundaries, although there is some suggestion that multidate principal components may be sensitive to vegetation boundaries, and texture to physiographic boundaries.

[1]  R. Lunetta,et al.  Remote sensing and Geographic Information System data integration: error sources and research issues , 1991 .

[2]  King-Sun Fu,et al.  A survey on image segmentation , 1981, Pattern Recognit..

[3]  J. Franklin,et al.  Land cover stratification using Landsat Thematic Mapper data in Sahelian and Sudanian woodland and wooded grassland , 1991 .

[4]  V. Judson Harward,et al.  Mapping forest vegetation using landsat TM imagery and a canopy reflectance model , 1994 .

[5]  J. L. Moigne,et al.  Refining image segmentation by integration of edge and region data , 1992, IEEE Trans. Geosci. Remote. Sens..

[6]  Russell G. Congalton,et al.  A practical look at the sources of confusion in error matrix generation , 1993 .

[7]  Curtis E. Woodcock,et al.  Nested-hierarchical scene models and image segmentation , 1992 .

[8]  D. C. Mason.  Segmentation of terrain images using textural and spectral characteristics , 1979 .

[9]  David C. Mason,et al.  Segmentation of remotely-sensed images by a split-and-merge process+ , 1988 .

[10]  K. Thapa,et al.  Accuracy of spatial data used in geographic information systems , 1992 .

[11]  J. Franklin,et al.  Testing the Woodcock-Harward image segmentation algorithm in an area of southern California chaparral and woodland vegetation , 1996 .

[12]  M. Gahegan,et al.  A model to support the integration of image understanding techniques within a GIS , 1996 .

[13]  Andrew K. Skidmore,et al.  Map accuracy assessment using line intersect sampling , 1992 .

[14]  Frank W. Davis,et al.  Improved integration of remote sensing and geographic information systems: a background to NCGIA initiative 12 , 1991 .

[15]  C. Woodcock,et al.  Combining Spectral and Texture Data in the Segmentation of Remotely Sensed Images , 1996 .

[16]  P. August,et al.  GPS for environmental applications: accuracy and precision of locational data , 1994 .

[17]  Jr. James Edward Frew.  The image processing workbench , 1991 .

[18]  J. Franklin,et al.  Coniferous Forest Classification and Inventory Using Landsat and Digital Terrain Data , 1986, IEEE Transactions on Geoscience and Remote Sensing.

[19]  G. Edwards,et al.  Modeling uncertainty in photointerpreted boundaries , 1996 .