Assessing Uncertainties in Landuse/Landcover Classification using Remote Sensing Images

• Main Authors: Lin-Hsuan Hsiao, 蕭淩瑄
• Other Authors: Ke-Sheng Cheng
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/50079900175914256053

碩士 === 國立臺灣大學 === 生物環境系統工程學研究所 === 101 === Multispectral remote sensing images are widely used for landuse/landcover (LULC) classification. Performance of such classification practices is normally evaluated through the confusion matrix which summarizes the producer’s and use’s accuracies and the overall accuracy. However, the confusion matrix is based on the classification results of a set of multi-class training data. As a result, the classification accuracies are heavily dependent on the representativeness of the training data set. It is imperative for practitioners to assess the uncertainties of LULC classification in order to obtain a full understanding of the classification results. In addition, the Gaussian-based maximum likelihood classifier (GMLC) is widely applied in many practices of LULC classification. The GMLC assumes the classification features jointly form a multivariate normal distribution, whereas, in reality, may features of individual landcover classes have been found to be non-Gaussian. Direct application of GMLC will certainly affect the classification results. In the study conducted in Taipei and its vicinity, the satellite images acquired by the AVNIR-2 sensor onboard the ALOS satellite were used. We tackled those two problems by firstly transforming the original training data set to a corresponding data set which forms a multivariate normal distribution before conducting classification using GMLC. Then, we applied the bootstrap resampling technique to generate a large set of multi-class resampled training data set from the original training data set. LULC classification was the implemented for each resampled training data set using GMLC. Finally, the uncertainties of LULC classification accuracies were assessed by evaluating the distributions of the accuracies derived from a set of confusion matrices. Combining the resampled bootstrap results of classification for each pixel and setting a threshold, pixels with higher uncertainties would be assigned to “unclassified”. The spatial characteristics of the uncertainties of LULC classification were assessed by showing the location of the unclassified pixels in the map. Results of this study demonstrate that Gaussian-transformation of the original training data achieved better classification accuracies, and that the bootstrap resampling technique is a very helpful tool for assessing uncertainties of LULC classification because it could assess the uncertainties in classification accuracies and illustrate the mixed pixels in the study area.