RESEARCH ON VISUALIZED DATA QUALITY CONTROL METHODS OF GROUND OBJECT SPECTRUM IN YANZHOU MINING AREA

摘要

Errors or outliers are prone to be made on account of various accidental factors or system errors in the observation process of ground object spectrums. It is necessary to carry on some rigorous gross error detection and quality control measures on field spectroscopy data before which is conducted to further spectral analysis. To this end, in this paper, in accordance with measured data of several typical crops in Yanzhou mining area, a theory of cluster analysis for field spectroscopy data quality controlling was proposed and 4 different cluster methods included Statistical distance, Aitchison distance, Pearson's correlation coefficient and Multidimensional Vector Cosine were used in the gross error visualized detection. For the common characteristic bands of different spectrum data, the goal of visualized detection and identification of outliers was achieved by means of the statistical method of box-and-whisker plots. Outliers which were identified can be getting rid of in the use of several self-developed graphic interactive controls based on GDI+ technology. The theory proposed in this paper provided effective quality assurance for in-depth spectroscopy analysis.