FILTER SELECTION FOR NIR SENSING OF PLANT AND SOIL MATERIALS

A method of selecting optical filters for use in near infrared sensors is presented. This method considers the collective effects of illumin ation source, reflectance characteristics of the surface or surfaces i n question, possible filter types and transmittance characteristics, a nd detector sensitivity. Reflectance ratios were generated for two or more detector/filter combinations and discriminant analysis was applie d to select the most meaningful complement of hardware. For illustrati on purposes this approach was applied to an example situation in which it was desired to discriminate between ''green plant matter'' and ''o ther'' materials. The ''other'' category was comprised of soil and cro p residue. A classification accuracy of 100% was possible with one ref lectance ratio generated from one detector type and two filters.