Automated Detection of Single Wheat Kernels Containing Live or Dead Insects Using Near–infrared Reflectance Spectroscopy

An automated near–infrared (NIR) reflectance system was used over a two–month storage period to detect single wheat kernels that contained live or dead internal rice weevils at various stages of growth. Correct classification of sound kernels plus kernels containing live pupae, large larvae, medium–sized larvae, and small larvae averaged 94%, 93%, 84%, and 63%, respectively. Pupae + large larvae calibrations were developed for live (day 1) and dead (days 7, 14, 28, 42, and 56) internal insects. Validation results showed that the live pupae +live large larvae calibration correctly classified 86% to 96% of dead pupae + dead large larvae validation samples. The dead pupae + dead large larvae calibration correctly detected the presence of live pupae + live large larvae with an accuracy of 92% to 93%. Thus, wheat kernels containing either live or dead insects can be used to develop calibrations for detecting both live and dead insects in wheat. These findings will impact how calibration sample sets can be handled. Results indicated that immediate sample processing for creating calibrations may no longer be necessary; internal insects can be killed and calibrations created at a later time without sacrificing accuracy. Additionally, laboratories can share these same calibration samples to save time and resources.