Elsevier

Biosystems Engineering

Volume 117, January 2014, Pages 104-113
Biosystems Engineering

Special Issue: Image Analysis in Agriculture
Research Paper
Internal characterisation of fresh agricultural products using traditional and ultrafast electron beam X-ray computed tomography imaging

https://doi.org/10.1016/j.biosystemseng.2013.07.002Get rights and content

Highlights

  • Ultrafast CT technology used to study fresh in vivo product components.

  • Hounsfield unit (HU) values obtained from CT images used to determine internal characteristics.

  • Information supplied to develop classification algorithms to sort fresh products.

  • CT technology is suggested as appropriate for inline sorting.

Currently, destructive techniques can be employed to evaluate the internal attributes of fresh fruits, vegetables and nuts. However, clearly not all produce can be evaluated. Thus, there is a need to develop an in vivo non-destructive technique able to assess fresh agricultural commodity internal components, especially disorders. In this study, medical grade computed tomography (CT) was used to obtain transversal two-dimensional (2D) images from several fresh agricultural product phenomena. CT scanning was performed by placing and securing numbered samples onto a whole polyethylene sheet, placed on the CT scanner table. Phenomena included the internal decay of chestnuts (Castanea spp.), internal defects in pickling cucumbers (Cucumis sativus), translucency disorder in pineapples (Ananas comosus), pit presence in tart cherries (Prunus cerasus var. Montmorency) and plum curculio (Conotrachelus nenuphar) infestation of tart cherries. In addition, an ultrafast X-ray CT scanner was also used to visualise internal characteristics of fresh chestnuts. Chestnuts were labelled and packed in a thin plastic hose, which was pulled through the scanning plane. The 2D CT X-ray images and post-processing three-dimensional CT image reconstruction indicate that CT can be used as an accurate in vivo insight of fresh intact agricultural products. Results suggest that there is a potential for non-destructive inline sorting of the internal quality of several agricultural products. The long-term objective is that the fresh and processing product industries will then be able to detect internal quality attributes of fresh agricultural commodities, at a relatively early stage, after validation under commercial conditions.

Introduction

It is estimated that a total of about 500 million tonne of fresh fruits, vegetables and nuts are produced annually in the world, representing a large, highly important and growing industry (FAO, 2010). At the moment, approximately 25–30% of the total production is discarded after harvest, mainly because of undetectable internal quality problems, safety issues and senescence (Kader, 2002). These represent large product and economic losses disturbing agriculture industry sustainability and the ability to consistently offer healthy, safe and good quality produce.

Fresh and processed agricultural commodity quality is measured not only by external factors such as colour, shape, size, surface blemish (gloss) and surface mold, as described in studies done by Jha and Matsuoka (2002) using fresh eggplants, but also by internal disorders and freshness, which are very important for consumer acceptance. Most importantly, the external appearance usually is not altered, at least initially, by internal disorders making them difficult to detect without destructive evaluation. Internal disorders usually are the result of anatomical and physiological changes within the tissue such as moisture loss, chemical conversion, discolouration, senescence, microorganism attack, cell breakdown (physiological decay) and insect injury (Upchurch et al., 1993). Physical and mechanical internal properties of commodities, which determine maturity, are also important for the overall quality of commodities (Jha, Kingsly, & Chopra, 2006).

In some cases these problems, including the lack of maturity, can lead to a completely unmarketable product. Additionally, negative responses from consumers can jeopardise the marketing of fresh produce causing severe economic losses. Because of the lack of in vivo diagnostic tools, little is known about how and when cell breakdown progresses and microorganisms infest fresh agricultural products (Butz, Hofmann, &Tauscher, 2005). Moreover, it is difficult to track pre- and post-harvest handling strategies that negatively or positively affect the quality and internal components of fresh and processed agricultural products. This includes, the effect of mechanical harvesting, storage (Jha et al., 2006), pre-harvest treatments (Mandujano et al., 1998, Sieber et al., 2007), peelability (Guyer, Fulbright, & Mandujano, 2005), optimal storage conditions, quality measurement and standards (Mencarelli, 2001), as well as in vivo morphology for cultivar characterisation (Ertan, 2007) and physiology (Jha et al., 2006).

Currently, destructive techniques can be employed to evaluate internal quality attributes, characteristics and components. However, this is not an in vivo technique and while trying to determine quality, not all commodities can be evaluated (Butz et al., 2005, Donis-Gonzalez et al., 2012a, Donis-González et al., 2012b, Donis-Gonzalez et al., 2010). In recent years, techniques based on two-dimensional (2D) X-ray and computed tomographic (CT) imaging have been explored and are used for non-destructive determination of internal characteristics of a variety of agricultural and food products (Abbott, 1999, Cubero et al., 2010, Donis-Gonzalez et al., 2012a, Donis-González et al., 2012b, Haff, 2008, Jha et al., 2010, Kotwaliwale et al., 2011). In addition, techniques based on optical, magnetic resonance imaging (MRI), near-infrared (NIR), vibration, sonic and ultrasonic, have also been explored, and are used for non-destructive determination of internal quality of a variety of agricultural and food products (Cubero et al., 2010, Lorente et al., 2011, Milczarek et al., 2009). Despite considerable research effort, in vivo component characterisation and real-time inspection systems of internal quality of fresh produce, are still uncommon in the industry, mainly because of limitations in useful information (e.g. implemented CT classification procedures) particularly when using high-speed systems (Butz et al., 2005). Additionally, because of the morphological characteristics of some commodities, for example chestnuts, which contain a thick external shiny shell, it is not possible to use traditional colour or even NIR sorting methods available in the industry to evaluate internal quality attributes. However, because of recent advances in high-performance computers, new detector technologies, the availability of advanced CT and traditional X-ray systems, high-performance X-ray tubes, real-time imaging, cost diminution and a significant decrease in image acquisition time, the field of non-medical CT applications and in-line CT sorting systems is becoming attractive (Hanke, Fuchs, & Uhlmann, 2008).

X-rays are short wave radiation (approx. 0.01–10 nm) with an energy between 1.92 × 10−17 and 1.92 × 10−14 J, which can penetrate matter. X-rays are generating by bombarding electrons on a metallic anode (X-ray generator) (Bushberg, Seibert, Leidholdt, & Boone, 2002). Traditional CT is an imaging procedure where an X-ray tube is rotated around an object or objects and the attenuation is recorded on a detector (Fig. 1a). Other equipment may contain a rotating stage in front of a fixed (non-moving) X-ray tube and detector (Bushberg et al., 2002).

Alternatively, newer and ultrafast techniques exist and are being further developed for potential fast inline imaging. Figure 1b shows the working principle of the ultrafast Rossendorf electron beam X-ray tomograph (Rossendorf Fast Electron beam X-ray tomograph – ROFEX) scanner. An electron beam of sufficient energy is produced by an electron beam gun, focused onto a semicircular X-ray production target, which surrounds the test section. The electron beam is swept rapidly across the target by means of an electromagnetic deflection system and X-rays are generated from a moving focal spot. Radiation passes through the object of investigation and is attenuated regarding to Lambert's law. Radiation intensity signals are recorded by a fast X-ray detector, which can capture images at a rate of up to 7000 frames s−1. The detector is designed as a circular ring and mounted stationary inside the scanner head with some axial distance to the plane of the focal spot path. From sets of projection data acquired from 240°, superimposed cross sectional images can be reconstructed using a conventional filtered back projection algorithm. The scanner is optimised with respect to ultrafast imaging, which implies a small source-detector separation and shorter electron path, accordingly. Objects with a diameter up to 80 mm can be scanned without any limited angle artefacts, which can be described as the loss of resolution in a 2D CT image because a limited number of available projection images (Bushberg et al., 2002, Fischer et al., 2008).

There are advantages of CT compared to traditional 2D X-ray imaging. First, CT completely eliminates the superimposition of images of structures within the samples and outside the region of interest. This is because in 2D X-ray imaging, only one projection image (X-ray transmission through sample) is acquired per sample, while in CT a transverse 2D-image or slice is reconstructed using information from more than one 2D projection image, acquired at different angles. Secondly, data from one CT imaging procedure can be reconditioned to be observed in various planes, or even observed volumetrically by creating a three-dimensional (3D) image, merging the information from several 2D slices. Finally, because of the intrinsic contrast and high resolution of CT, differences between materials that diverge in physical density by about 0.5% can be differentiated (Bushberg et al., 2002). In CT the difference in physical density of materials is visualised by changes in image intensity and it is expressed in ‘Hounsfield-Units’ (HU) (or ‘CT-number’). Hounsfield-Units represent the X-ray attenuation capabilities of a specific material. In a 2D CT X-ray image, the HU(x,y) in each pixel (x,y), of the image is generated through a combination of X-ray projection images and by using Eq. (1),HU(x,y)=1000μ(x,y)μwaterμwaterwhere μ(x,y), is the floating point number of the (x,y)-pixel before slice reconstruction, μwater is the attenuation coefficient of water (approximately 0.195) and the HU(x,y) is the Hounsfield unit observed in the final reconstructed CT-image. Therefore, objects with a low-density, such as air (at standard temperature and pressure), have a low HU (−1000 HU) and high-density materials, like bone, will have a high HU (up to 3000 HU). In general, a HU-value equal to 0 stands for the density of water (100.0 kg m3); values in the positive range represent materials with a mass density above 100.0 kg m−3; and values in the negative range stand for those below 100.0 kg m−3 (Bushberg et al., 2002).

CT methods for accurate visualisation, segmentation and inner component identification of fresh agricultural commodities, which include the internal decay of chestnuts (Castanea spp.), internal defects in pickling cucumbers (Cucumis sativus), translucency disorder in pineapples (Ananas comosus), pit presence in tart cherries (Prunus cerasus var. ‘Montmorency’) and plum curculio (Conotrachelus nenuphar) infestation of tart cherries, are not available. Therefore, we hypothesise that CT scanning can be used as an in vivo tool to accurately visualise and study the internal attributes/characteristics, known as regions of interest (ROIs), of these fresh agricultural products. For that reason, the objective of this study is to scan these fresh agricultural products, using a traditional and an ultrafast CT scanner, to evaluate if their ROIs can be visualised and validate the technology beyond visualisation by determining if HU-values from these ROIs are statistically different.

Section snippets

Sample preparation, in vivo two-dimensional computed tomography imaging scanning and three-dimensional image reconstruction

Samples of physiologically mature chestnuts (n = 400, year = 2009 and 2010), cucumbers (n = 80, year = 2009) and tart cherries (n = 250, year = 2009 and 2012) were obtained directly from commercial farms in Michigan, USA. Physiologically mature pineapples (n = 55, year = 2011) were acquired from commercial producers in Costa Rica and then shipped to Michigan State University – Department of Biosystems and Agricultural Engineering. After receiving samples, chestnuts, cucumbers and tart cherries

Results

Figures 3a–6a include examples of cross-sectional 2D CT images of uncut fresh chestnuts (Fig. 3a), tart cherries (Fig. 4a), pineapples (Fig. 5a) and pickling cucumbers (Fig. 6a). In these images, healthy and rotten chestnuts, curculio damaged cherries, translucent pineapples and physiologically affected pickling cucumbers can be viewed.

Parallel colour images of fresh slices, which correspond to approximately the same CT scanned slices, can also be observed for each commodity, as noted in

Discussion

Visually, using CT images, differences in HU-values (grey scale changes in 2D CT images) can be distinguished between healthy and rotten chestnuts, healthy tart cherries versus those containing pits or that are damaged by insects, translucent versus healthy pineapples and healthy against physiologically defective pickling cucumbers, by observing differences in intensities in the image, resulting from different HU values. In certain cases (e.g. pineapple translucency), because of dealing with

Conclusions

The medical grade traditional GE-CT imaging system provided 2D CT images of the internal structure, issues and components of fresh agricultural commodities. Significant changes and variation of HU-values associated with changes in internal characteristics (ROIs) of chestnuts, tart cherries, pineapples and pickling cucumbers could be discerned. In addition, an ultrafast ROFEX limited-angle-type CT provided 2D CT images of the internal structure of fresh chestnuts. Furthermore, reconstructed 3D

Acknowledgements

This work was financially supported by the Ernie and Mabel Rogers Endowment and Project GREEEN at Michigan State University, USA. The authors acknowledge the support of Professor Dennis Fulbright, Mr. Mario Mandujano from the Department of Plant, Soil and Microbial Sciences and Mr. James Burns from the Department of Biosystems and Agricultural Engineering at Michigan State University, for their valuable support and help in obtaining and preparing samples. We thank Mr. Mark Sellers, Mr. Rex

References (30)

  • E.G. Barcelon et al.

    Relating x-ray absorption and some quality characteristics of mango fruit (Mangifera indica L.)

    Journal of Agricultural and Food Chemistry

    (1999)
  • J.T. Bushberg et al.

    The essential physics of medical imaging

    (2002)
  • P. Butz et al.

    Recent developments in noninvasive techniques for fresh fruit and vegetable internal quality analysis

    Journal of Food Science

    (2005)
  • S. Cubero et al.

    Advances in machine vision applications for automatic inspection and quality evaluation of fruits and vegetables

    Food and Bioprocess Technology

    (2010)
  • I.R. Donis-Gonzalez et al.

    Shell mold and kernel decay of fresh chestnuts in Michigan

    Acta Horticulturae

    (2010)
  • Cited by (51)

    • Emerging non-destructive imaging techniques for fruit damage detection: Image processing and analysis

      2022, Trends in Food Science and Technology
      Citation Excerpt :

      This technology has a competitive edge since the X-rays can pass through most of the objects. The X-rays fall within the wavelength range of 0.01–10 nm, with a frequency and energy range of 30–30000 Peta Hz and 0.12–120 keV, respectively (Donis-Gonzalez, Guyer, Pease, & Barthel, 2014). Specifically, X-rays having wavelength ranging from 0.1 to 10 nm are known as soft X-rays with energy varying between 0.12 and 12 keV, which are commonly used for food inspection (Cai, Yuan, Liu, & Sun, 2014).

    • Sensor coating employed to preliminarily evaluate the banana ripeness

      2021, Colloids and Surfaces A: Physicochemical and Engineering Aspects
      Citation Excerpt :

      To verify the adhesion of the film to the substrate, we constructed 3-D images of bananas based on 3-D reconstruction (Fig. S2). CT scan is a nondestructive testing technique for the diagnosis of internal diseases in fresh agricultural products [29]. CT hounsfield units (HUs)of fruit fractions, film, and air are different.

    View all citing articles on Scopus

    Developed from a presentation at the IV International Workshop on Computer Image Analysis in Agriculture, held at CIGR-AgEng 2012, Valencia, Spain: July 8th–12th, 2012.

    View full text