Multispectral Bands Selection of Horticultural Plant Diseases Using Brightness

We persent a new method for performing bands selection experiments with 16 narrow-bands multispectal images data of cucumber. This method allows for the visual inspection base on the brightness value of the images which achieve standard assessment during visual wavelength region. It follows a standard feature selection approach in which correlative coefficient measure is used as a figure of merit in a search-optimization procedure. For this paper, we have chosen 6 cucumber leaf's Pseudoperonospora samples with compound symptom . Brightness value of multispetral images is used over 16 bands for the visual inspection. 4 bands are precluded. Then correlative coefficients of the rest 12 multispectral images are calculated. 530nm, 546nm, 650nm, 670nm and 850nm , 5 bands are selected by their correlative coefficients. The result shows that the method is independent of observer of interest base on brightness of images. And it is appropriate to inspecting patterns. It insures that all information contained in original features is preserved and guarantees to make use of data class information. It also implies the possibility to design new devices to perform measurements only for few spectral regions that make sense for discrimination.