| The launch of the first commercial, high-resolution remote sensing satellite, IKONOS, in September 1999 marked a revolution in the remote sensing industry, Capable of 4m multispectral and 1m panchromatic images at a four-day revisit rate. IKONOS quickly became a household name. Awed by its potential, industries and governments at all levels clamored to incorporate these new products and services into their daily business. Among other applications municipalities adopted IKONOS as an effective means of monitoring change in residential and commercial construction, roads and green-space. However, this practice has been hampered by the lack of high-resolution imagery before September 1999.;The purpose of this report is to examine the extent to which such localized change detections can be performed using high and medium-resolution satellite imagery. To this end a post-classification change detection was completed using 4m IKONOS data and 30m Landsat 5 data. Since the accuracy of the change detection algorithm is dependent on the compounded accuracies of the two component classifications the selection of an appropriate classifier is of the utmost importance. As such, each image was classified using the maximum likelihood, IsoData and neural network classification schemes. Based on the results of these classifications it was determined that the neural network classifier was most suitable for both data sets. The final result suggests that urban change can be effectively monitored using medium and high-resolution imagery through the application of post-classification comparison and a well-trained neural network classifier. |
