Assessment of the area measurement on Cartosat-1 image

The goal of this study was the evaluation of agriculture parcel area measurement accuracy on Cartosat-1 imagery, and the determination of the technical tolerance appropriate for measurement using photointerpretation techniques. A further objective was to find out the influence of image type, land cover or parcel size on the area measurement variability. In our experiment, five independent operators measured 185 parcels, 3 times, on each image. Next, the buffer width, calculated as the difference between measured and reference parcel area, was derived and was the subject of statistical analysis. Prior to verifying the normality of the buffer widths, a detection of anomalous measurements is recommended. This detection of outliers within each group of observations (i.e. parcels) was made using the Jacknife distance test on each type of imagery (Cartosat Aft, Cartosat Fore). Then, the General Linear Model procedure to identify major significant effects and interactions was followed by analysis of variance to ease the interpretation of the variability observed of the area measurement. Finally, two different parameters, reproducibility limit and critical difference, were calculated to make comparison with other sensors like digital aerial orthophoto in this study possible. The repeatability limits gave the acceptability difference between two operators when measuring the same parcel. For orthophoto this value reached 2.86m, on Cartosat-1 5.17m and 8.76m for Aft and Fore image respectively.