Detection of spongy pulp in guava using light properties and near infrared spectroscopy

Abstract

Spongy pulp defect in guava is recognised by a dry-looking pulp with a brown colour, and not acceptable to consumers which causes substantial loss in value. Current detection of spongy pulp uses visual assessment of the flesh, which is half cut from the sample. The present work aimed to develop a classifying model based on non-destructive technique for the detection of spongy guava. Guava samples harvested at full maturity were determined for visible light properties, visible light reflectance, and near infrared reflectance. The light properties and light reflectance of guava peel were used to derive a classification model which was then compared with a near infrared reflectance model, which in turn provided absorbance of the flesh and peel using stepwise discriminant analysis. The models were used to classify the guavas into normal and spongy flesh groups, which were assigned with reference to the visual assessment on half cut samples. The classification accuracy for the model using gloss and light reflectance at 650 nm (chlorophyll b) was 90.4%. However, the model developed from the near infrared absorbance provided better accuracy (92.7%). It appeared that the largest wavenumber at 4,721 cm-1 contributed to the total sugar content, which implied that spongy and normal guavas had different total sugar contents in the flesh. The present work demonstrated the potential of near infrared spectroscopy to discriminate spongy from normal guavas. However, the accuracy of the classification could be further improved by analysing more samples from the next season.