Rapid Determination of Starch in Fresh Cassava by Near-Infrared Spectrometry

A NIRS method for use in field or rapid analysis of fresh cassava for its starch content was proposed. Ninety samples of various species were taken and their contents of starch were determined by the chemical method given in GB-standard. NIR spectra of these samples in shape of cyclindrical cubes were obtained by scanning through their corss section, and transformed into their 1st derivative graphs. Mathematical model for NIRS was established by using the QPLS pathern cube of the CAUNIFT software. It was shown that linear correlationship was obtained between the values of starch contents found by the chemical method and the predicted values obtained by the math. model. Accuracy and feasibility of the math. model were checked further by analyzing 5 unknown samples, and as shown by the results, absolute errors of within 2% and relative crrors of less than 5% were found between the values of starch contents (w%) found by chemical method and the predicted values obtained from the NIRS math. model.