Biometrical Approaches for Mandarin Leaf Image Analysis

A study examining correlations between fruit trees physiological and visual properties collected by remote sensing technology started at Budapest University of Economic Sciences and Public Administration, Faculty of Horticultural Science, together with Kyoto University and Wakayama Research Center of Agriculture, Forestry and Fisheries, Fruit Tree Experiment Station (Japan). Experimental plant was Satsuma mandarin (Citrus unshiu MARC. var. satsuma), the physiological property to examine was leaf water potential, which correlates with the sugar content of mandarin fruit according to preceding studies. Large amount of visual information had been recorded at ground level and was processed by the methods of image analysis and biometry. Under field conditions images of given leaves showed different type of reflectance frequency distribution. These distributions often could be identified as normal distribution, however left skewed, right skewed and bimodal distributions also occurred even partially shifted out of the perceptive range of the measuring device. Comparison of the applied biometric methods was analyzed to adjust with the natural field conditions, hardware and software features. Results showed that normal distribution model could be applied for analyzing data obtained at optimal conditions, however these phenomena under natural field conditions could often be described better when weighted averaging methods were applied on frequency distributions.