Application Of Robust Estimation In The Surveying Adjustment

Surveying adjustment actually is facing three errors, including accidental errors, system errors and gross errors. Generally the traditional surveying adjustment only considers accidental errors because all kinds of specifications have been formulated for the original observation data. When the raw data do not meet these specifications, it is necessary to survey them again. Therefore, we can think that the original data don't include gross errors and system errors and there are only accidental errors in the surveying adjustment. Because of its excellent statistical properties in dealing with accidental error, least-squares method has been used as the basis of surveying adjustment for several hundred years, and its theoretical study and practical application have been made a very rapid progress.But with the development of surveying science and technology, in particular the global positioning system and remote sensing applications, data collection has been made more and more modern, fast, automated, large-scale. It is impossible to use traditional methods to rule out gross errors, so consideration for the accidental errors and gross errors simultaneously has become a research direction in the surveying circles, in the early 1980s, Robust estimation was introduced into the surveying circles, because of its obvious advantages in the processing gross errors and accidental errors, it has been extensive researched and rapid developed. At present, the Robust estimation has become a major method in handling errors, but any method has its deficiencies. In the application of Robust estimation, some problems still exist, mainly:(1) how to solve the problem that ture errors are replaced by correction number.(2) how to choose the initial weight is also a problem.(3) related observation Robust estimation theory needs further study.This paper tries a little superficial research on these issues and provides some simple ideas, mainly:(1) provide a two-steps surveying adjustment approach to improve the coefficient matrix of normal equation and observation component matrix, making the ture errors to distribute into the corresponding correction number as much as possible after the adjustment, reduce the problems caused by replacement true errors with correction number in the weight-selected iteration .(2) under different circumstances we should use different method including least squares initial value or the one norm minimum estimation initial value.(3) summarize the various estimation related to the observation methods, including the robust estimation methods and other relevant observations estimation methods.Finally, use a leveling network surveying adjustment to prove that these arguments are feasible.