Based on hyperspectral cameras from the Finnish manufacturer Specim, Researchers from Wageningen University & Research (WUR) have developed a smart All-in-one Spectral Imaging (ASI) laboratory system for standardized, automated data acquisition and real-time spectral model deployment. A fully standalone system integrated with two different hyperspectral cameras working in the complementary spectral ranges has proven its capability for reliable fresh produce analysis.

Hyperspectral imaging has been used in scientific and industrial applications for quite a while. In numerous use cases, hyperspectral imaging has proven to be a powerful technology for advanced classification and material analysis. Compared to conventional imaging systems, hyperspectral imaging extends beyond the visible spectrum, making it capable of detecting information that is invisible to the human eye.

Hyperspectral imaging offers unparalleled capabilities in numerous fields where precise characterization and identification are essential. While hyperspectral imaging has demonstrated significant advantages and potential in various industries and research fields, the complexity of data acquisition and analysis has hindered its widespread adoption – up until now.

To overcome this challenge, Dr. Puneet Mishra and his colleagues from WUR in The Netherlands developed a system that enables users without in-depth knowledge of hyperspectral imaging to benefit from this technology. Hyperspectral line-scan cameras from the Finnish manufacturer Specim have been the critical component of the project.

In the traditional image processing market, numerous systems, such as smart cameras, enable users to solve tasks without machine vision knowledge. There is no such push-button type of imaging system for hyperspectral imaging, and there was a need to build an easy-to-use one-touch system. Mishra et al. (2022) developed an intelligent All-in-one Spectral Imaging (ASI) laboratory system for standardized, automated data acquisition and real-time spectral model deployment to simplify hyperspectral image acquisition.

Fig. 1: Wageningen University & Research´s smart All-in-one Spectral Imaging (ASI) laboratory system for standardized automated data acquisition and real-time spectral model deployment. Image source: Wageningen University & Research

Standardized hyperspectral imaging

Several hurdles had to be overcome in the past to realize practical implementations of spectral imaging for routine analysis. One of the primary challenges with spectral imaging is that the available cameras on the market usually require system integration and calibration modeling.

In addition, most of the hyperspectral sensors in the market are currently supplied as data acquisition tools. For conducting the measurements and developing the model, the user needs to design experimental setups where the sensor must be integrated. The best practice followed in research laboratories is performing the data acquisition and modeling in separate steps. Although this approach has shown potential, non-experts cannot consider it a practical solution for routine use.

To make hyperspectral imaging easier, especially for non-expert users, WUR researchers developed a standardized spectral imaging system with embedded computing to have minimal influence on the measurement from unwanted sources. The goal was to avoid system reintegration and to reuse calibrated models to allow repeatable measurement.

Fig. 2: The ASI system has proven its reliability by predicting properties in several fruits. Image source: Wageningen University & Research

Hyperspectral cameras as key components

The critical decision for the successful realization of the system was the selection of suitable hyperspectral cameras. On this point, the developers at WUR opted for hyperspectral line-scan Specim FX10 and FX17 cameras from Specim. For the system’s desired compact design, the Specim cameras’ mechanical size was perfect. In terms of resolution and speed, these cameras were also able to meet the specifications without any problems.

Complementing the Specim cameras, researchers integrated a controlled, standardized illumination environment, an in-built computing system, as embedded software for automated image acquisition and model deployment. This setup was the foundation for exploring the spatial distribution of sample properties in real time in the ASI system.

Fig. 3: All-in-One spectral cabinet with Specim FX10 and FX17 cameras. The data modeling is inspired by complementary data fusion and performed with special data fusion methods developed by Dr. Puneet Mishra. Image source: Wageningen University & Research

Combining VNIR and NIR cameras for more precise inspection

To show the capability of the ASI framework, WUR researchers conducted exemplary cases of fruit property analysis. ASI setup capabilities to analyze fruit property were demonstrated for wide fruit cases such as grapes, cherry, pear, and kiwi. The human eye and traditional machine vision cameras are sensitive to wavelengths between 380 and 760 nm. Being limited to that range, knowing the chemical parameters related to fruit maturity, such as moisture and soluble solids content, is difficult. In contrast, the extended wavelength range of hyperspectral cameras makes it possible to predict parameters such as moisture and soluble solids content.

The machine initially used a Specim FX10 hyperspectral camera that operates in the visible and near-infrared (VNIR) spectral range from 400 to 1000 nm. With only FX10, it was possible to check the sugar content and some other features of the fruits but to realize equipment for higher requirements to collect more detailed information with higher quality that would be capable of reliably analyzing, for example, meat and other food an additional InGaAs-based Specim FX17 near-infrared (NIR) hyperspectral camera that covers the wavelength range from 900 to 1700 nm was added. The sensor fusion opened further options for analyzing food and other organic objects.

Fig. 4: Specim’s FX10 and FX17 hyperspectral cameras are the critical components for the fruit inspection system, covering a wavelength range from 400 to 1700 nm. Image source: Specim

Specim FX10 and Specim FX17 complement each other, especially when measuring in high-moisture samples such as fresh fruit. In the case of fresh fruit, it has been found that due to low water absorption coefficients of water molecules in the 400 – 1000 nm spectral range, the penetration depth of the VNIR light is higher, which allows for capturing more subsurface bulk information. In the 900 to 1700 nm range, the water absorption coefficient of the water molecule is high, thus allowing a better analysis of surface moisture in samples.

The Specim FX17 camera is flexible regarding the recording speed as it offers the option of selecting and evaluating from 224 wavelength bands and only using those that supply relevant information for the current application. By reducing the number of observed wavelengths, the standard Specim FX17 recording speed of 670 lines per second when using all 224 wavelength bands can be increased to several thousands of lines per second when focusing on just a few wavelength bands. This property is called Multi Region of Interest (MROI). It is available on both the Specim FX10 and FX17 cameras and gives users very high flexibility in terms of speed without losing accuracy. In addition, MROI reduces the data quantity for easier processing and data storage. Using Specim FX10 and FX17 cameras in one system improved inspection precision and enabled quality examinations in the full wavelength range from 400 to 1700 nm.

Fig. 5: Using Specim’s FX10 and FX17 hyperspectral cameras together allowed a more precise prediction of fruit property than using only one camera. Image source: Specim

Reliable fresh produce analysis with hyperspectral imaging

The All-in-one spectral imaging system developed by Dr. Puneet Mishra and his colleagues is based on Specim FX10 and FX17 hyperspectral cameras and has met the requirements to precisely analyze moisture and soluble solids content in a range of fresh fruits. The system was benchmarked in performance against commercial point spectrometer systems that are widely used for NIR analysis. The ASI system achieved similar performance as the long-term established technology, and there were only insignificant differences between the prediction of the ASI setup and the commercial spectrometers. In addition, hyperspectral imaging offers other advantages over point spectrometers.

A key benefit of the ASI development compared to point spectrometers is that it allows exploring spatially distributed properties due to the rich spatial information captured by the Specim FX hyperspectral cameras. In addition, the ASI setup enables the reuse of existing spectral data and models that have been acquired before during laboratory experiments. This opens up new options for broader usage of spectral sensing where models and data can be shared between different spectroscopy users.

Furthermore, ASI is a fully mobile system that can be brought to the samples instead of bringing the samples to the lab. In many use cases, this is a huge advantage over traditional technologies. Last but not least, it takes less than 40 seconds to get the results. In the past, one might have to wait several days for the results of moisture content analysis.

Combined with its ease of use, the ASI system gives even non-expert users a chance to access the potential of this technology. Researchers are already using the ASI system for experiments with all kinds of food products. Due to the device’s portability, the system’s possibilities are far from exhausted.

Fig. 6: Overview of the All-In-One spectral imaging (ASI) setup and workflow for Kiwi fruit analysis. Image source: Wageningen University & Research

Fig. 7: Dr. Puneet Mishra: “The ASI system opens up new options for a wider usage of spectral sensing where models and data can be shared between different users of spectroscopy, and it can even be used by users with only a little knowledge in the field of Spectral Imaging.”
Image source: Wageningen University & Research

About Specim:

Specim is a globally leading supplier of hyperspectral imaging and a true pioneer and forerunner in this field. An international team of more than 90 professionals with expertise in optics, electronics, software, and machine vision serves the market with the broadest range of hyperspectral cameras, imaging spectrographs, systems, and accessories. Specim is a trusted partner for industrial OEMs, machine builders, and integrators.

References

Amigo, J. M., Babamoradi, H., & Elcoroaristizabal, S. (2015). Hyperspectral image analysis. A tutorial. Analytica Chimica Acta, 896, 34-51. https://doi.org/https://doi.org/10.1016/j.aca.2015.09.030

Lu, Y., Saeys, W., Kim, M., Peng, Y., & Lu, R. (2020). Hyperspectral imaging technology for quality and safety evaluation of horticultural products: A review and celebration of the past 20-year progress. Postharvest Biology and Technology, 170, 111318. https://doi.org/https://doi.org/10.1016/j.postharvbio.2020.111318

Mishra, P., Sytsma, M., Chauhan, A., Polder, G., & Pekkeriet, E. (2022). All-in-one: A spectral imaging laboratory system for standardised automated image acquisition and real-time spectral model deployment. Analytica Chimica Acta, 1190, 339235. https://doi.org/https://doi.org/10.1016/j.aca.2021.339235

Mishra, P., Verschoor, J., Vries, M. N.-d., Polder, G., & Boer, M. P. (2023). Portable near-infrared spectral imaging combining deep learning and chemometrics for dry matter and soluble solids prediction in intact kiwifruit. Infrared Physics & Technology, 131, 104677. https://doi.org/https://doi.org/10.1016/j.infrared.2023.104677

Mishra, P., & Xu, J. (2023). Multimodal close range hyperspectral imaging combined with multiblock sequential predictive modelling for fresh produce analysis. Journal of Near Infrared Spectroscopy, 31(3), 141-149.

Related products:

Specim FX10

Specim FX17