Analyzing stone fruits ripeness: Linking sugar content to absorbance spectra

Stone fruits characterized by the large single seed in their center (also known as the stone or pit), stone fruits include apricots, cherries, raspberries, olives, and even coconuts.. It's important to harvest these fruits when they are at their peak ripeness since they won't ripen after being picked. When buying stone fruits, customers can check for ripeness by squeezing (a hard fruit indicates that it is not fully ripe) or smelling the fruit (the more aromatic the fruit, the riper it is) .

Although this could work for a consumer purchasing a few pieces of fruit, how can distributors sort fruits so that their shipments contain only the best, ripest fruits? Since the concept of ripeness is somewhat arbitrary and challenging to measure, it is frequently associated with a fruit's increased sugar content with ripening. A refractometer is a useful tool for measuring sugar content. The unit Brix, which is defined as 1 g of sucrose in 100 g of solution, is easy to use to determine sugar concentration. Unfortunately, extracting juice from each fruit is necessary for this sugar content measurement, which is detrimental because the fruit is no longer fit for sale to customers. For commercial operations who are interested in characterizing and sorting fruits of various types by the ton, non-contact near infrared spectroscopy can be used to measure brix accurately and efficiently.

Distinguishing fruits that are suitable for harvest is done based on reflectance-absorption measurements. The light source used can provide a Vis-Nir light range in the region from 450 nm to 1100 nm. The non-contact testing technique aims to predict the sugar content and ripeness of fruit, while also keeping the fruit intact for delivery to consumers. Ocean Insight (Ocean Optics) offers sturdy, reliable Vis-NIR Light Source (such as Krypton Light Source and Tungsten Halogen Light Sources) and Vis-NIR spectrometers (Ocean SR2 VIS-NIR, Ocean SR4 VIS-NIR, Ocean SR6 VIS-NIR, Ocean  HR2 VIS-NIR, Ocean HR4 VIS-NIR, Ocean HR6 VIS-NIR, Ocean FX-VIS-NIR, Ocean-HDX-VIS-NIR) that are suitable for this technique.

For example, test on different peaches with different degrees of ripeness. The peaches were stored in a refrigerator to prolong the timeframe for their ripeness but were left in a room-temperature environment for a minimum of four hours before measurements were taken to ensure the sample temperature had minimal effect of the recorded data. Each sample was individually placed on the turntable center and allowed to spin freely under the overhead lamp. 10 measurements of each peach sample were taken to quantify the average spectrum of each sample, create more data points for the algorithm creation, and determine if any measured spectra were outliers that could skew the algorithm. Each peach sample was then sliced in two opposite positions on the fruit, and the fruit of each slice was individually squeezed onto the refractometer window to determine the Brix value of each slice. The two Brix values for each peach sample were averaged. Because the spectra used were absorbance spectra, need to do reference measurement with white reference tile at a similar distance and angle as the peach samples. Ocean Insight (Ocean Optics) has researched and developed reflectance sample, which is ideal  for this reference measurement.

With measured brix values as pricing attributes. After selecting the default values for the spectral range for calibration and the number of coefficients for analysis and calibration, the initial algorithm was created. The input data was adjusted to remove any outlier spectra to provide the most accurate prediction algorithm based on measurements of the first excavation samples. Finally, additional peach samples were measured with a refractometer, comparing the algorithm's predicted sugar content values against those measured with the refractometer.

After removing outlier spectra, the total spectra were used to create the updated algorithm. The spectra of the measurements do not fall outside the values shown in Figs. This updated algorithm produced the following results for predicted sugar content versus actual sugar content in 10 measured samples

Ultimately, this updated algorithm provides a much closer linear fit to the data set.

ANALYSIS AND CONCLUSION

After removal of what were considered outlier spectra, the algorithm very closely predicted the sugar content of each peach sample. With the algorithm closely predicting the sugar content of the peach samples used to create it, additional measurements were taken of the four other peach samples to verify the accuracy of the algorithm. The same method was used, with ten absorbance measurements taken of each peach sample followed by two slices cut from each peach on opposite sides of the fruit to determine actual sugar content with the refractometer. 10 of the results are given in the table below:

The results show a high degree of accuracy in predicting the sugar content of peach samples 9 and 10 but are much less accurate for samples 11 and 12. This could be due to the fact that samples 11 and 12 were beginning to show signs of rotting, including slightly wrinkled skin, whereas the inner juice from the fruit slice may still contain the same level of sugar. Alternatively, this may be an indication that creating such a predictive model requires a larger sample size and additional pieces of equipment to take even more scans of the fruit.

CONCLUSION

To sum up, the current experiment demonstrates how absorbance measurements can be used to forecast the sugar content and, consequently, the ripeness of peach samples. More advanced techniques would need to be employed to determine which spectra are regarded as valid or outliers. It should be noted that even the four samples utilized to assess the accuracy of our algorithm did contain outlier spectrum that were eliminated. A genuine average spectrum for each fruit is determined by using many fiber optics and spectrometers to gather large amounts of data, as fruit sorting is typically an in-line process with fruit moving swiftly. Although this straightforward experiment showed the technique's potential, substantial data collection and modeling are needed to create reliable and accurate models that can be used in a commercial environment for high speed fruit sorting.