Detection of Pecan Weevil Larvae in Pecan Nutmeat Using Multispectral Imaging System

Abstract

This project utilizes multispectral imaging techniques to detect and identify pecan weevil larvae in pecan nutmeat. Diffuse reflectance measurements were obtained for weevil larvae and pecan nutmeat using a VIS/NIR spectrometer. PCA and derivative analysis were performed to identify the spectral wavelengths that best differentiated pecan nutmeat from larvae. The four potential wavelengths in the spectral range of silicon CCD were identified as 855nm, 902nm, 940nm and 981 nm. The images were acquired with a NIR enhanced camera. The images acquired at 980nm showed significant gray scale contrast between pecan nutmeat and larvae. These images were then processed using masking and morphological processing. This method was compared to a novel active contour based image segmentation algorithm. The contour based algorithm produced much better segmentation results and should be used instead of simple masking operation. Classification accuracy of 84% was obtained for the training images and 74% for the testing images.