Peach Variety Detection Using Vis-Nir Spectroscopy

INTRODUCTION

Accurately identifying peach varieties is critical to ensuring agricultural quality and meeting market demands. Traditional methods for identifying many peach species such as DNA analysis, gas chromatography, and amino acid composition, are often labor-intensive, time-consuming, and expensive. In order to overcome these limitations, VIS-NIR (visible and near-infrared) spectroscopy was proven to be a powerful tool. The technique measures the reflectance of light in the visible and near-infrared ranges, providing detailed information about the internal composition and properties of a sample in a non-destructive manner. When combined with advanced data analysis methods, VIS-NIR spectroscopy can quickly and accurately classify fruit varieties based on their unique spectral signatures. This application note explores how different peach varieties can be detected using a VIS-NIR spectrometer, highlighting its potential to revolutionize quality control in the agricultural sector.

METHOD

To predict different types of peach, a model was established by using VIS-NIR spectral variables as an input, and the final output was the prediction results of peach varieties. VIS-NIR (visible and near-infrared) spectroscopy is a powerful analytical technique that measures the light reflectance of a sample in the visible and near-infrared regions. The model trained using spectral data produced highly accurate results.The training dataset consisted of 350 peach samples, the first test dataset consisted of 75 peach samples, and the second dataset consisted of 75 peach samples, with the same number in each species. Training and validation datasets are used to optimize the model. By comparing the test dataset and the train dataset, the model can classify different types of peach. The chart of the training and test process are shown in Fig 1.

Figure 1 The chart of training and test process

The results of five different peach samples are shown in Figure 2, where the red curve represents the spectrum of "Zao Yanhong" peach, the green curve represents the spectrum of "Zaofengwang" peach, the purple curve for the spectrum of "Taqiao" peach, the olive curve is the spectrum of "Baifeng" peach, and the black curve for the spectrum of "Yin Shuanghong" peach.

Figure 2 VIS-NIR spectrum of peach samples

Regarding training accuracy, the model achieved 100% accuracy on the training dataset. This shows that the model is able to perfectly classify peach varieties based on the trained spectral data. Regarding test accuracy, the model maintained a high accuracy of 94.4% when tested on an independent dataset. This demonstrates the robustness of the model and its ability to generalize better than expected to new unknown samples. These results highlight the effectiveness of using VIS-NIR spectroscopy combined with a prediction model to classify peach varieties.

SYSTEM

The VIS-NIR spectra were collected in diffuse reflectance mode using a small VIS-NIR spectrometer (STS, Ocean Insight) equipped with an optical fiber probe, a tungsten halogen lamp HL-2000-FHSA and optical fibers.

Ocean Insight's STS micro-spectrometer is a compact and powerful device. Its small size and rugged construction make it ideal for integration into a variety of systems, providing high optical resolution, sensitivity, and fast data acquisition rates. It is compatible with OceanView software, providing a user-friendly experience for applications such as environmental monitoring, quality control, and scientific research. The portability and versatility of the STS-VIS make it a valuable tool for precise spectral analysis in a variety of environments.

Figure 3 STS spectrometer

The HL-2000-FHSA is a high-powered halogen light source designed for spectroscopic applications. Key features include a 20-watt tungsten halogen lamp with an integrated attenuator for controlling light output, and a TTL-shutter for automated control. The light source covers a broad spectral range from 360 to 2400 nm and is suitable for use in VIS-NIR spectroscopy. It provides stable, high-intensity illumination, making it ideal for applications such as reflectance and absorbance measurements.

Figure 4 HL-2000-LL-FHSA Light Source

CONCLUSION

To sum up everything that has been stated so far, integrating VIS-NIR spectroscopy into a predictive model provides a highly accurate, non-destructive, and efficient method for detecting peach varieties. Moreover, the high training and testing accuracies achieved in the study demonstrate the robustness and reliability of the method. Furthermore, this approach significantly reduces the time and effort required for traditional detection techniques, making it a valuable tool for the agricultural industry. By quickly and accurately identifying peach varieties, VIS-NIR spectroscopy combined with data analysis methods improves quality control processes, supports decision-making for peach retailers, processors, and consumers.