Theory, Application And Program Design Of Quasi-Accurate Detection

Based on the principle of Quasi-Accurate Detection (QUAD),the implementation scheme,applications and program design of the method are systematically studied in this thesis.Firstly,the basic theory of the QUAD is introduced,then the implementation scheme is discussed in detail. The key of this method is how to select the quasi-accurate observations (QAOs) correctly,so the indicators of the QAOs selection are researched in details in this thesis. A successful scheme is that the appropriate QAOs are chosen through two steps including preliminary and fine selection. For the prior information is not known in practice,the preliminary selection of QAOs is mainly based on the standardization residuals. However,the effect of gross errors on the standardization residuals are influenced by observation space and the data structure,thus the observations are divided into four types to some specific criteria and the QAOS are chosen by the value of standardization residuals and the classification information of the observations. The estimators of real errors would appear a "hive off' phenomenon if there existed gross errors and the preliminary selection of the QAOs are correct. Fine selections of QAOs are based on the "hive off phenomenon of real error estimators. The observation estimators that the real errors are smaller than the given threshold values are selected as QAOs until the "hive off' characteristic of real error estimators does not change. With all the gross errors removed from observations by using QUAD method,theparameter estimators are relatively reliable and the precision is relatively high.The QUAD method was put forward to detect gross errors at first,but in this thesis,the QUAD method is extend to the fields of block division,deformation detection,circle slip detection in rapid GPS positioning and resolution of ill-conditioned and polluted equation. Its efficiency is better than that of the hypothesis method and some robust schemes especially for the instance that the observations are severely polluted. For example,in block division,the pollution rate is larger than 50%,the hypothesis method and robust method are helpless but the QUAD method can successfully divided the block into several parts with different trend each by reasonable selection of QAOs. The QUAD method can detect n-m-1 gross errors in theory,if there exist,(n is the number of observations,m is the number of unknown parameters) and its breakdown point is much high,so its application prospect is very broad in many fields.The QUAD method program is preliminarily designed with MATLAB language for its powerful matrix function,simple language and friendly interface. In order to keep secret of the code sources,increase run speed and use without the MATLAB environment,the program is compiled to EXE file so as to be called freely by other computer language. The automatization of program is realized to help users using it conveniently without manual intervention.