36 - 
relative comparisons of spectral and spatial classification performance 
should be valid. The results from training areas are encouraging and indi 
cate that for the categories investigated, a sorting of image areas into 
Level I categories may be performed. If this sorting is done using spatial 
features, the method could be implemented by optically scanning an ERTS frame 
in only one band, possibly using an on-board satellite system (Akram, 1973). 
Detailed classification of individual pixels into Level II or high levels 
could then be attempted using either spectral or spatial information. 
ACKNOWLEDGEMENTS 
This work was supported by the Earth Observations Division of the 
Johnson Spacecraft Center, NASA under Contract NAS9-12972. The authors would 
also like to express their appreciation to Mr. George Hartinger, Colorado 
State University, for computer assistance. 
REFERENCES 
Akram, S. 
1973: 
Design concepts for an on-board parallel image processor 
Pattern Recognition, Pergamon Press, Voi. 5. p. 3-11 
Anderson, J. E. Hardy, and J. Roach. 
1973: A land-use classification system for use with remote-sensor 
data. Geological Survey Circular 671, U. S. Department of 
Interior. 16 p. 
Gramenopoulos, N. 
1973: Automated thematic mapping and change detection of 
ITEK Corp., Goddard Space Flight Center Report. 63 
Haralick, R., K. Shanmugam and I. 
1973: Textural features for 
Man Cybernetics. Vol. 
Di nstein. 
image classification. IEEE 
SMC-3, No. 6. p. 610-621 
ERTS-A images. 
P- 
Trans. Systems, 
Hornung, R. and J. Smith 
1974: Incorporation of spatial information as additional features for 
multispectral land-use classification. Final Report, Vol. I 
Contract NAS9-12972. Colorado State University, Fort Collins, 
Colorado. 179 p. 
Kirvida, L. 
1973: Automatic photointerpretation for plant species and stress 
identification (ERTS-A). Final Report, Contract NAS5-21866. 
Honeywell, Inc., St. Paul, Minn. 60 p.