Research Of Data Processing Method With Incomplete Observations Base On EM Algorithm

In surveying, due to a variety of unfavorable factors, the measurement data often contains gross errors or miss data. The measurement data become incomplete data. Conventional adjustment methods are base on complete data. When the missing data is essential to the estimation of the parameter, if the missing data is not considered, the accuracy of the estimation will be seriously affected. Nowadays, the handling methods for missing data mainly are imputation, fitting, prediction and so on. Using the measures which is used to process the missing data can effectively improve the quality of the data processing, as well as measurement accuracy and reliability. However, all those methods in need of fulfilling the data, then the estimation base on the complete data. The problem is that the fitting data is inaccurate data. Using it must be affecting the accuracy of the estimation. In another fact, fitting data is unnecessary in the estimation of the geomatics. Therefore, this thesis has done some research in processing methods of the incomplete observations base on EM (expectation-maximization) algorithm. Mainly about:1. In the paper, we introduced and contrasted the commonly used incomplete data processing measures, analyzed the advantages and disadvantages of the various measures, searched for the suitable measure to the incomplete measurement data.2. EM algorithm is applied to the process the incomplete measurement data. Combined with the conventional processing model, derived the formula of the data processing. Illustrated the processing steps using EM algorithm, realized the incomplete measurement data processing method based on EM algorithm.3. Promoted the application of the EM algorithm in the measurement data processing. EM algorithm is applied to the process the measurement data with uncertain information. Derived the formula of the data processing, illustrated the processing steps using EM algorithm, and realized the measurement data processing method with uncertain information based on EM algorithm, improve the accuracy of the measurement results. 4. Using concrete examples to prove the feasibility and effectiveness of the incomplete measurement data processing method and measurement data processing method with uncertain information based on EM algorithm.