Objective  A high-efficiency, real-time, inline assay using near infrared spectroscopy to determine the moisture content in coconut meat was developed.

  Method  The near infrared (NIR) spectra of 360 mature copra specimens came from varied germplasms were scanned by a grating diffuse NIR instrument made in China (S400 NIR Spectrophotometer for Quality of Agricultural Products). The collected data were separated by half into modeling and test sets. The quantitative partial least squares analysis was applied for the model construction with data pretreatments, such as first derivative, second derivative, scatter correction, zero-centered, and range normalization.

  Result  The developed models achieved optimal performance. The root mean square error of calibration (SEC) and that of prediction (SEP) on the models were 0.7605 and 0.8378, respectively, while the correlation coefficients on the modeling and test sets 0.9963 and 0.9960, respectively.

  Conclusion  The newly developed method for the determination of copra moisture content was capable of instantly and precisely delivering the measurement. It could be employed for real-time detection on a coconut processing line for quality assurance and yield prediction. Moreover, similar approach could conceivably be used to develop protein, fat, and carbohydrate determinations for copra as well.