Application of thermal imaging combined with machine learning for detecting the deterioration of the cassava root.

Freshness is an important parameter that is indexed in the quality assessment of commercial cassava tubers. Cassava tubers that are not fresh have reduced starch content. Therefore, in this study, we aimed to develop a new approach to detect cassava root deterioration levels using thermal imaging with machine learning (ML). An underlying assumption was that nonfresh cassava roots may have fermentation inside that causes a difference in the inner temperature of the tuber. This creates the opportunity for the deterioration level to be measured using thermal imaging. The features (pixel intensity and temperature) that were extracted from the region of interest (ROI) in the form of tuber thermal images were analyzed with ML. Linear discriminant analysis (LDA), k-nearest neighbor (kNN), support vector machine (SVM), decision tree, and ensemble classifiers were applied to establish the optimal classification modeling algorithms. The highest accuracy model was developed from thermal images of cassava roots captured in a darkroom under a control temperature of 25 °C in the measurement chamber. The LDA, SVM, and ensemble classifiers gave the best overall performance for the discrimination of cassava root deterioration levels, with an accuracy of 86.7%. Interestingly, under uncontrolled environmental conditions, the combination of thermal imaging plus ML gave results that were of lower accuracy but still acceptable. Thus, our work revealed that thermal imaging coupled with ML was a promising method for the nondestructive evaluation of cassava root deterioration levels.

Modified specific gravity method for estimation of starch content and dry matter in cassava.

An empirical model for the estimation of starch content (SC) and dry matter (DM) in cassava tubers was developed as an alternative method to polarimetry and dry oven. These improved estimation equations were developed based on the specific gravity (SG) method. To improve accuracy, the one hundred-seventy-four sample were obtained from four commercial varieties of cassava in Thailand including KU50, CMR38-125-77, RY9 and RY11, respectively. The age of sample collected from four to twelve months after planting was used in this experiment. The empirical model was created from their relationships between SG obtained from small sample size (~100 g) and its SC and DM. The SG for cassava was strongly correlated with the SC and DM, with values for the coefficient of determination (R2) of 0.81 and 0.83, respectively. The SC showed a high correlation with the DM, with R2 of 0.96. To confirm that the empirical model was effective when applied to other samples, unknown samples collected from another area were tested, and the results showed a standard error of prediction (SEP) of 1.02%FW and 3.49%, mean different (MD) of -0.66%FW, -0.89% for the SC and DM, respectively. Hence, our empirical equation based on a modified SG method could be used to estimate the SC and DM in cassava tubers. It can help breeders to reduce costs and time requirements. Moreover, breeders could be used the methods to evaluate the SC and DM from the tuber formation to harvesting stage and monitoring the changes in SC and DM during breeding.

Predicting Marian Plum Fruit Quality without Environmental Condition Impact by Handheld Visible-Near-Infrared Spectroscopy.

Handheld near-infrared spectroscopy was used to study the effect of integration time and wavelength selection on predicting marian plum quality including soluble solids content (SSC), the potential of hydrogen ion (pH), and titratable acidity (TA). For measurements representing actual conditions, the on-tree fruits were scanned under in-field conditions. The assumption was that the robust model might be achieved when the models were developed under actual conditions. The results of the main effect test show that the integration time did not statistically affect SSC, pH, and TA predictions (p-value > 0.05) and the wavelength range had a significant impact on prediction (p-value < 0.01). An integration time of 30 ms coupled with a wavelength range of 670-1000 nm was the optimal conditions for the SSC prediction, while an integration time of 20 ms with 670-1000 nm wavelength was optimal for pH and TA prediction because of the lowest root-mean-square error of cross-validation (RMSECV). The optimal models for SSC, pH, and TA could be improved using spectral pre-processing of multiplicative scatter correction. The effective models for SSC, pH, and TA improved and reported the coefficients of determination (r 2) and root-mean-square errors of prediction (RMSEP) of 0.66 and 0.86 °Brix; 0.79 and 0.15; and 0.71 and 1.91%, respectively. The SSC, pH, and TA models could be applied for quality assurance. These models benefit the orchardist for on-tree measurement before harvesting.