Research On Directional Data Of Measurement

The data used to express angle or directional random experimental result is called directional data. Firstly, this paper introduced some related knowledge on directional data, pointed the directional data existing in surveying and proved that the directional data such as angle observation and their errors are von Mises normal distribution rather than Gauss normal distribution, which was used beforetime. Thisconclusion was verified with the test of goodness of fit to aim directly at a set identical triangulation data.Up to now, the angle observations in surveying is thought as Gauss normal distribution and was processed with least square method in classic adjustment. However, it was proved that they are directional data and are von Mises normal distribution rather than Gauss normal distribution. So the observation and errors of them aren't satisfying the hypothesis of LS estimation, thus in classic adjustment method there are some defects here. In this paper Box-Cox transformation is used to make the observation errors of directional data satisfy the hypothesis of LS estimation. An applied example was given and proved the new method was more reasonable. The new method was used in practical project and proved the new method was better than classical method. Finally, some conclusion and suggestions of following research work were given.