All of the Landsat 7 data collected after 2003 contains missing pixels in the form of unsightly stripes across the images. To recover missing data of a Landsat image, different methods may be used. However, the gap fi...All of the Landsat 7 data collected after 2003 contains missing pixels in the form of unsightly stripes across the images. To recover missing data of a Landsat image, different methods may be used. However, the gap filling process creates inconsistencies on pixel intensity values. The incongruous pixel numbers are anomolous observations and their classification in the reference specter is challenging. In an effort to contribute to this need, we propose a reliable robust approach to classify inconsistent pixels after the gap filling process. To estimate multivariate location-scale parameters a new robust DMVV （depth minimum vector variance estimator） is presented. The DMVV algorithm does not require any matrix inversion for its calculation, consequently its computational time is highly reduced. The results show that it has a high breakdown point and is very efficient for large data set. Landsat remote sensing data of Jakarta Province across years 2002 and 2010 are used as case study.展开更多
文摘All of the Landsat 7 data collected after 2003 contains missing pixels in the form of unsightly stripes across the images. To recover missing data of a Landsat image, different methods may be used. However, the gap filling process creates inconsistencies on pixel intensity values. The incongruous pixel numbers are anomolous observations and their classification in the reference specter is challenging. In an effort to contribute to this need, we propose a reliable robust approach to classify inconsistent pixels after the gap filling process. To estimate multivariate location-scale parameters a new robust DMVV （depth minimum vector variance estimator） is presented. The DMVV algorithm does not require any matrix inversion for its calculation, consequently its computational time is highly reduced. The results show that it has a high breakdown point and is very efficient for large data set. Landsat remote sensing data of Jakarta Province across years 2002 and 2010 are used as case study.
