A liquid crystal tunable filter based shortwave infrared spectral imaging system: Design and integration
Highlights
► A liquid crystal tunable filter (LCTF) based shortwave infrared (SWIR) spectral imaging system. ► The system mainly consisted of an LCTF, an InGaAs camera, a SWIR lens, and an illumination unit. ► The system can capture hyperspectral or multispectral images in the spectral range of 900–1700 nm. ► The methodologies for designing the LCTF-based spectral imaging system were introduced in detail.
Introduction
Spectral imaging (SI) is a rapidly growing area for nondestructive inspections of food safety and quality. It often refers to hyperspectral imaging (HSI), which acquires images at hundreds of contiguous spectral bands. Spectral imaging that uses a small number of discrete spectral bands (<10) is often called multispectral imaging (MSI) (Lu and Park, 2008). In the last decade, SI has gained recognition for food safety and quality inspections due to its nondestructive inspection capabilities. It captures images of a test object at a number of narrow bandpass wavelengths. As a result, spectral images contain a wealth of spatial and spectral information of the test object, which can be used to investigate the intrinsic physical/chemical properties of the test object (Lu and Chen, 1998). The sensing capabilities of SI have been demonstrated by numerous applications reported in a broad range of food and agricultural products such as poultry (Park et al., 2010), beef (Naganathan et al., 2008), wheat (Singh et al., 2010), apple (ElMasry et al., 2008, Kim et al., 2008), citrus (Qin et al., 2009), and cucumber (Ariana and Lu, 2009).
Two technologies are commonly used by SI systems for food quality and safety inspection (Gowen et al., 2007): (i) imaging spectrograph based technology which disperses light with wavelengths across the CCD detector using a prism-grating-prism element resulting in a 3-D image cube obtained by line scanning; and (ii) electronically tunable filter (ETF) based technology which selects wavelength bands, takes spatial 2-D images one wavelength at a time, and then reconstructs the data into a 3-D image cube. Liquid crystal tunable filters (LCTF) and acousto-optical tunable filters (AOTF) are the two most commonly used ETFs. The LCTF-based HSI systems are relatively slower than the systems using AOTFs, but the former have better imaging performance since the LCTFs have larger apertures, relatively wider field of view, and lower wavefront distortions than the AOTFs (Evans et al., 1998).
In hyperspectral imaging, the line-scan technique has been predominantly used in food safety and quality inspection due to its high speed and high spectral resolution. In addition, compared to the ETF-based HSI technique, the line-scan technique is easier to integrate with the current online inspection systems using conveyor belts. Nevertheless, the ETF-based HSI system has its own advantages over the line-scan system. First, an ETF-based hyperspectral imaging system is a natural extension of a multispectral imaging system, which provides versatility to the system to be used for either HSI or MSI applications. Second, an ETF-based HSI system has an area of field of view (FOV), whereas the line-scan HSI systems can only see one line of the test object at a time (Kerekes and Schott, 2007). Third, the ETF-based HSI systems can select spectral bands randomly in addition to continuously (Gat, 2000). Thus, the ETF-based HSI systems have the advantage in instantaneous imaging applications requiring selective spectral information. Fourth, the parameter setting of an ETF-based SI system, such as the exposure time of the camera, is often dynamic and adjustable over each spectral band in a scan, while a line-scan system often has to keep its parameter setting constant during scanning. Another important advantage of the ETF-based HSI systems is that they do not rely on moving mechanical devices such as a linear conveyer module (Sun, 2010). Therefore, they are more compact, can be easily integrated with other applications, and have higher potential for field application.
Developing an ETF-based spectral imaging system requires many special considerations on system design, integration, and calibration. For instance, the low optical throughput of the LCTF requires special considerations with the selection of the lens and the design of the lighting. Gat (2000) summarized the fundamental principles and considerations of integrating an spectral imaging system using an ETF, with an example of integrating an LCTF-based HSI system in the range of 400–1000 nm. However, the paper did not discuss detailed methodologies for designing and calibrating the ETF-based spectral imaging system. Gebhart et al. (2007) quantitatively characterized their LCTF-based fluorescence and diffuse reflectance imaging system (400–720 nm) for detecting brain tumor demarcation. But methods and processes for designing and calibrating the system were less discussed in the article.
In food safety and quality inspection, generally, the ETF-based HSI systems have not been fully studied and used. In early applications, Evans et al. (1998) demonstrated an LCTF-based spectral imaging system (650–1050 nm) for studying plant health. Archibald et al. (1999) used a similar system to differentiate color class of wheat kernels. Later, Cogdill et al. (2004) reported an LCTF-based HSI system for predicting the concentrations of maize kernels’ constituents by using the hyperspectral transmittance imaging. Gomez-Sanchis et al. (2008) used an LCTF-based HSI system to detect the rottenness in mandarins in the spectral range of 460–1020 nm. Recently, a number of LCTF-based HSI applications in shortwave infrared (SWIR) range were reported. Williams et al. (2009) evaluated the hardness of the maize kernels by using both an LCTF-based (960–1662 nm) and a line-scan (1000–2498 nm) HSI systems. Similar LCTF-based SWIR spectral imaging systems (960–1700 nm) were applied to differentiate Canadian wheat classes (Mahesh et al., 2008) and to identify insect-damaged wheat kernels (Singh et al., 2009). However, most of the reported studies focused on presenting their applications, whereas the design and integration of their systems were seldom discussed.
This article demonstrates the methodology of designing an LCTF-based spectral imaging system in the spectral range of 900–1700 nm. The specific objectives were to: (1) present the methodology used in the design and integration of an LCTF-based SWIR spectral imaging system; (2) validate the system by differentiating four different food materials.
Section snippets
The principle of the LCTF-based spectral imager
An LCTF-based spectral imager is basically a conventional imaging system with an extra electronic band-pass filter. An LCTF, essentially, is a multistage Lyot–Ohman type polarization interference filter using a stack of polarizers and tunable retardation liquid crystal plates (Tran, 2005). The filter element of the LCTF filters light according to the choices of polarizers, the retarder, and the liquid crystal waveplate. Since the retardance of each filter element is adjustable, a narrow
System architecture
Fig. 6 shows the architecture of this LCTF-based SI system. The system was designed on a multi-tier architecture so that functional modules were logically separated and reusable. Five independent logical tiers, including hardware (2 tiers) and software (3 tiers), were designed. The bottom tier included the spectral imager and the illumination system. The next tier consisted of a computer and two I/O hardware interfaces which connected the spectral imager and the computer. The top three tiers
Demonstration
Four common agricultural materials, sugar, wheat flour, water, and 95% ethanol, were scanned by this LCTF-based SWIR system to demonstrate the system’s capabilities in distinguishing agricultural materials. The goal of the test was to differentiate two liquid materials (water and ethanol), and to differentiate two solid materials (wheat flour and sugar) in both spectral and spatial domains. In the test, 3 ml of 95% ethanol and 3 ml of water, and 5 g of pure sugar and 5 g of wheat flour were placed
Conclusions
This paper presents the methodology of designing and integrating an LCTF-based SWIR spectral imaging system. In particular, special considerations were discussed in detail on how to select individual components of the system (lens, detector, and illumination unit). In addition, the optimal configuration of the system was determined by considering the layout of the imager and optimizing the configuration of the lighting. The efficacy of our LCTF-based SWIR spectral imaging system was
Acknowledgements
This work was funded by the United States Department of Agriculture National Institute of Food and Agriculture Specialty Crop Research Initiative (Award No. 2009-51181-06010), Georgia Food Industry Partnership, and Vidalia Onion Committee. The authors also gratefully acknowledge Mr. Gary Burnham and Mr. Tim Rutland for their assistance in developing this system.
References (32)
- et al.
Early detection of apple bruises on different background colors using hyperspectral imaging
LWT-Food Science and Technology
(2008) - et al.
Hyperspectral system for early detection of rottenness caused by penicillium digitatum in mandarins
Journal of Food Engineering
(2008) - et al.
Hyperspectral imaging-an emerging process analytical tool for food quality and safety control
Trends in Food Science and Technology
(2007) - et al.
Feasibility of near-infrared hyperspectral imaging to differentiate canadian wheat classes
Biosystems Engineering
(2008) - et al.
Visible/near-infrared hyperspectral imaging for beef tenderness prediction
Computers and Electronics in Agriculture
(2008) - et al.
Detection of citrus canker using hyperspectral reflectance imaging with spectral information divergence
Journal of Food Engineering
(2009) - et al.
Detection of insect-damaged wheat kernels using near-infrared hyperspectral imaging
Journal of Stored Products Research
(2009) - et al.
Identification of insect-damaged wheat kernels using short-wave near-infrared hyperspectral and digital colour imaging
Computers and Electronics in Agriculture
(2010) - et al.
Maize kernel hardness classification by near infrared (NIR) hyperspectral imaging and multivariate data analysis
Analytica Chimica Acta
(2009) - et al.
Development of short-wavelength near-infrared spectral imaging for grain color classification
Proceedings of SPIE
(1999)
Evaluation of internal defect and surface color of whole pickles using hyperspectral imaging
Journal of Food Engineering
Current and future trends in infrared focal plane array technology
Proceedings of SPIE
Single-kernel maize analysis by near-infrared hyperspectral imaging
Transactions of the American Society of Agricultural Engineers
Development of a spectral imaging system based on a liquid crystal tunable filter
Transactions of the American Society of Agricultural Engineers
Imaging spectroscopy using tunable filters: A review
Proceedings of SPIE
Cited by (40)
Hyperspectral imaging for non-invasive blood oxygen saturation assessment
2024, Photodiagnosis and Photodynamic TherapyFully convolutional networks for blueberry bruising and calyx segmentation using hyperspectral transmittance imaging
2020, Biosystems EngineeringCitation Excerpt :The details of the handling methods for the bruised and unbruised fruit have been described by Zhang, Li, Takeda, et al. (2017). A liquid-crystal-tuneable-filter (LCTF) based near-infrared hyperspectral imaging (HSI) system (Wang, Li, Tollner, Rains, & Gitaitis, 2012; Zhang, Li, & Yang, 2017) was implemented to acquire transmittance images. The hyperspectral imager consists of an LCTF (Model Varispec LNIR 20-HC-20, Cambridge Research & Instrumentation, Cambridge, MA, USA), an InGaAs camera (320 × 256 pixels, 25 μm pitch, Model SU320KTS-1.7RT, Goodrich, Sensors Unlimited, Inc, Princeton, NJ, USA), and a lens (50 mm, F/1.4, Model SOLO 50, Goodrich, Sensors Unlimited, Inc, Princeton, NJ, USA).
Detection of blueberry internal bruising over time using NIR hyperspectral reflectance imaging with optimum wavelengths
2017, Postharvest Biology and TechnologyCitation Excerpt :An NIR hyperspectral imaging system has been set up in our laboratory and employed for measuring hyperspectral reflectance images of blueberries (Fig. 1a). The main components of the system are an LCTF (Model Varispec LNIR 20HC20, Cambridge Research & Instrumentation, MA, USA), an indium gallium arsenide (InGaAs) camera (320 × 256 pixels, 25 μm pitch, Model SUI320KTS1.7RT, GOODRICH, Sensors Unlimited, Inc, NJ, USA) with a 12-bit digital Camera Link compatible output, and an optical focusing lens (Nikkor 50 mm f/1.4D AF, Nikon, Japan) (Wang et al., 2012b). The illumination unit consisted of two 20-W halogen lamps adjusted manually at an angle of about 45° to illuminate the camera’s field of view.
Classification of foreign matter embedded inside cotton lint using short wave infrared (SWIR) hyperspectral transmittance imaging
2017, Computers and Electronics in AgricultureCitation Excerpt :The HSI consists of an LCTF (LNIR 20-HC-20, Cambridge Research &Instrumentation, Cambridge, MA, USA) and an indium gallium arsenide (InGaAs) SWIR camera (SU320KTS-1.7RT, Goodrich, Sensors Unlimited, Inc., Princeton, NJ, USA) combined with a near infrared lens (SOLO 50, Goodrich, Sensors Unlimited, Inc., Princeton, NJ, USA). The imaging procedure was controlled by a LabVIEW program—built in-house—using a computer (Intel® Pentium® D Processor, 4 GB DDR3, Windows 7) via Camera Link (Wang et al., 2012b) (Fig. 3). To obtain transmittance images, the sample was sandwiched between two floated borosilicate glass plates (BOROFLOAT® 33, thickness = 2.00 mm, Home Tech SCHOTT North America, Inc., Louisville, KY, USA) on the objective table above the illumination source.
Shortwave infrared hyperspectral reflectance imaging for cotton foreign matter classification
2016, Computers and Electronics in AgricultureCitation Excerpt :In addition, previous studies mostly used line-scan or push-broom scanning which required the object to move in a direction perpendicular to scanning lines so that the entire object can be imaged (Lu and Chen, 1998). Liquid crystal tunable filter (LCTF) based hyperspectral imaging system is an area-scan technique that captures spatial 2-D images at selected wavelengths sequentially and then stacks these images from different wavelengths together to form a 3-D hyperspectral image cube (Wang et al., 2012b). Compared to the line scan technique, the LCTF system does not require a moving sample and thus is especially suitable for developing a multispectral imaging system that can randomly access user-selected wavelengths.