Applying Remote Sensing Image Classification Techniques and Area Estimation Methods for Land Use Purposes

In this paper, two main techniques of area estimation for land use derived from remotely sensed data have been discussed. The first is conventional technique which refers to pixel counting estimator. The other technique is based confusion matrix area estimators including direct estimator, inverse estimator, additive estimator, bias removal technique and map marginal proportion based estimator. To evaluate the performance of these area estimator techniques a comparative study has been done using three evaluation criteria quantitative error, bias and dispersion. The results show that, all areas which are estimated by based confusion matrix area estimators are more accurate and closer to the true areas than that by conventional technique. Direct and additive estimators produce accurate and consistent results. The map marginal proportion based estimator and inverse estimator also produce accurate results in cases of large size testing samples.