Outlier Detection in Cadastral Coordinate Transformation Using Least Absolute Deviation

• Main Authors: Lin, Yi Chun, 林怡君
• Other Authors: Lin, Lao Sheng
• Format: Others
• Language: zh-TW
• Online Access: http://ndltd.ncl.edu.tw/handle/39414759906421467212

碩士 === 國立政治大學 === 地政研究所 === 101 === There are two coordinate systems with different geodetic datum in Taiwan region, i.e., TWD67 （Taiwan Datum 1967） and TWD97 （Taiwan Datum 1997）. In order to maintain the consistency of cadastral coordinates, it is necessary to transform one coordinate system to another. However, no matter what transformation method was used, the accuracy of the result is highly depended on the data quality. Since the uncertainty about whether outliers exist or not, so the outlier detection of data becomes an important work before coordinate transformation. The LAD（Least Absolute Deviation） method was affected by nearly very little or none from outliers. Thus, this method has been successfully used for outlier measurements detection in other fields. Therefore, LAD method was used to detect outliers in cadastral coordinate transformation in this study. This method provides an examination to ensure the quality of cadastral coordinates before converting one coordinate system to another. So, the accuracy of coordinate transformation can be increased. Then, the coordinate transformation results of LAD method and LS （Least Squares） method in the aspects of outlier detection ability and the accuracy of coordinate transformation were compared. On one hand, three varied sizes of simulating test areas, which contains different magnitude of outliers, at numbers of reference points and check points were placed, for checking the robustness and outlier detection ability in LAD and LS methods. On the other hand, data of real test areas were also used. Then, results from different coordinate transformation and adjustment methods were compared and analyzed, by using 4 and 6 parameter coordinate transformation respectively. The test results show that LAD method is more robust than LS method, and outliers can be detected easily from the residuals of reference points. While LS method is affected by outliers and the outlier detection ability is weaker than LS method. However, if the data contain none outliers, the coordinate transformation results by using LS method is better than LAD method. Therefore, it is suggested to using LAD method firstly. Then, after deleting all the outliers, one can use LS method to calculate coordinate transformation parameters.