Detection of locusts using near-infrared spectroscopy and cluster analysis

The main objective of this study, was to investigate the feasibility of the Near Infrared spectroscopy (NIR) techniques for rapid detection of locusts from their environment such as vegetation, using a Fourier Tran form Near-Infrared spectrometer (FT-NIR). Stones, grass, soils and locusts composed of all the 79 samples. The spectra of all samples were analyzed within a range of 3996-12489.5 cm(-1). Complex-linkage models were developed. The classification accuracy was calculated to assess the classification capability of calibration models. Different samples were correctly classified in rates of >91.67%%. The study demonstrated that using the established calibration models, locusts could be predicted with classification accuracy over 91.67%.