Advanced computational methods for spatial information extraction

A variety of mathematical approaches for spatial information extraction using digitized aerial photography and satellite imagery have been developed and implemented on serial computers. However, because of data volume and scale, the computational demands of spatial analysis procedures frequently exceed the capacity of available serial processing technologies. One way of addressing this problem is through parallel processing in which the power of multiple computing units can be used on a single problem. In this study we investigate the utility of parallel processing for spatial feature extraction. Our testing in the situation of texture feature extraction using a cooccurrence matrix indicates that dramatic reductions in execution time are possible-an image that required about 34 min to process using one processor was solved in under 2 min using nineteen processors. The availability of additional processors could result in smaller execution times. This speedup potential is a critical element in future studies focusing on more complex spatial analysis procedures. 1995.

Advanced computational methods for spatial information extraction Academic Article