Terahertz Non-Destructive Evaluation of Wine Corks and Eonology
Terahertz Non-Destructive Evaluation of Wine Corks
Natural cork is acquired from the Cork Oak (Quercus suber) predominately in Portugal and other countries surrounding the Mediterranean Sea. It is utilized in a variety of products including cork stoppers for wine and other beverages. As an enclosure for liquids, it has the desirable properties of being nearly impermeable to liquids and gases, as well as compressible. The quality of cork enclosures is determined by the presence and size of defects, voids, or cracks. Methods of NDE of corks include visible inspection of visible cracks and defects either by human experts or camera systems, chemical analysis (including cork soaks) for the presence of Trichloroanisole (TCA), and x-ray tomography.
The Federici research group examined the prospects of using Terahertz (THz) and millimeter wave (MMW) spectroscopy and imaging as a non-destructive evaluation tool of natural cork enclosures. Since the cork cells contain mostly gas, they are fairly transparent to THz radiation particularly in the MMW and low frequency THz rang thereby enabling imaging of voids, defects, as well as the cork's grain structure. Analysis of the spectral properties of the cork show that the THz absorbance can be modeled by a power law. A simple application of Mie scattering in the low scattering region qualitatively fits the observed spectral properties. The contrast in the THz transmission through different grains is likely due to variations in the cell structure that occur during the seasonal growth of the cork tree.
In the figure below, front and back visible (left) and THz (right) images through a 4.35mm (upper) sample from section of a wine cork. The THz image is formed by calculating the average transmitted power in the 0.9-1.0THz frequency band. The sample holder, which is made of aluminum, appears dark in the image since it blocks THz radiation. The sample is oriented so that the grain structure of the cork ies roughly parallel to the bottom of the page. Note the presence of cracks, void, and cork grain that are visible in the THz image that otherwise would be overlooked in examining only the front visible image.
While natural cork is nearly impermeable to liquids and gases, liquids do permeate the first ~ 2mm of a wine cork when it is stoppering a wine bottle. The cork swells due to the liquid which helps seal the bottle enabing the wine to slowly age. The defects and cracks in the cork enable a much more rapid diffusion of liquids through the cork structure. Below is illustrated an experiment which uses THz NDE to image water diffusion through the cylindrical sides of a wine cork. Visible images of cork cross sections for (a) front and (b) back surfaces. The back surface image is flipped horizontally so that the composite image(c) can be used to visualize the composite structure of the two surfaces in transmission. (d) illustration of the cuts for the circular (dotted lines) and rectangular (dashed lines) cork samples relative to the growth directions of the cork oak. (e) A schematic of the sample enclosure. The water diffuses into the cork sample perpendicular to the direction of THz propagation. In the visible images, the darker areas which a horizontal to the page are lenticels which permit the exchange of gases between the atmosphere and the internal structure of the truee.
The figure below shows the THz absorbance (0.65-0.7THz) through cork cross-section at 0hr (dry cork), 10.9hr, 21.9hr, 33hr, 44hr, 55.6hr, 78.2hr, and 93.6hr, respectively. Dark regions correspond to low absorbance while bright regions correspond to high absorbance. Bright regions outside of the cork are highly transparent in the dry image since the water was not be added to the sample chamber. Dark regions near the top of the sample chamber result from the level of water dropping in the chamber due to evaporation. If one compares the THz transmission images with the visible images of the cork disk, it is apparent that the water diffuses more quickly in the lenticels and voids.
Terahertz Imaging of Grapes - Crop Yield Estimation
The fundamental information needed to estimate the yield of any crop is the number of plant parts being harvested and the mass of each part. The concept of yield monitoring has been a part of precision agriculture.Yield prediction has been explored, along with pre-harvest efforts to map quality zones and thereby improve harvest efficiency and the quality of processed products. Pre-harvest estimates of yield and quality also have significant potential economic benefits by improving harvest efficiency and optimizing harvesting and processing . There is a growing interest in our ability to predict yield more accurately and at an earlier stage of development. Information derived from such predictions could be used to improve the operating efficiency of processing facilities, such as wineries, packing and cold storage facilities for table grapes, and grape concentrate plants.
Remote sensing utilizing satellite or airborne imaging in several spectral bands has been applied to agriculture monitoring. Airborne based millimeter wave (MMW) imaging has been used to measure the temperature and moisture/ ice content of soils . Typically by examining the difference in light reflection at two or more visible or near-infrared spectral bands, estimations of foliage health can be used to estimate the crop yield. For example, the Normalized Difference Vegetation Index (NDVI) uses spectral data from 580-680nm (corresponding to absorption from chlorophyll) and 725-1100nm (corresponding to high reflectance from the leaf structure) to create images of plant growth, vegetation cover and biomass production. In the case of crop estimation in vineyards, there is not a good correlation between multispectral satellite imaging and the crop yield; grape vines may be pruned several times in a growing season so that there is no strict correlation between the ‘greenness’ of the foliage or canopy size and yield. Moreover, since the grape clusters are typically partially covered by the canopy, direct visible imaging of the grape berries is difficult.
Wineries have a significant interest in predicting yield to plan harvest schedules and insure that adequate personnel, equipment, chemicals and storage space are available. In the event of insufficient yield, wineries may underestimate their ability to meet market demands, purchase excessive chemicals, supplies (bottles, cork, label, barrels, tank space) and labor. Excess yield has the opposite effect and can also result in significant losses in wine and wine quality due to an inability to process fruit in a timely manner while unpicked fruit is loosing quality or rotting. Regardless of which of these occur, it is generally accepted that our inability to accurately (+/- 5%) predict wine grape yield is a multimillion dollar problem within the United States alone. Since the US represents only a small percentage of world production, this makes yield prediction of wine grapes a significant economic problem internationally. Although growers generally have procedures they use to approximate or affect yield (pruning, thinning, manual cluster count), seasonal variability can significantly alter yield through changes in fruit set thereby changing the number of berries per cluster and changes in berry size. The latter is generally heavily influenced by the amount of available moisture just after fruit set and from veraison to harvest. These variables could therefore require a grower to make multiple estimates of yield and consequently require significant time.
The current practice for estimating grapevine yield involves the random selection of a few vines within a vineyard and manually counting the clusters. This is typically done early in the growing season since it is easier to see the clusters at this time. However, it is difficult to determine the number of berries per cluster and to predict the berry and therefore cluster weight at harvest. It is generally accepted that the contribution of these components (clusters per vine, berries per cluster and berry weight) to the final yield are approximately 60%; 30% and 10% respectively. Because the number of vines selected for counting is typically less than 1% it is difficult to accurately estimate a major contributor (60%) to the final yield.
In our earlier work on crop yield, we demonstred the capability of Terahertz imaging to sense the presence of grapes, their size, and the location hidden behind the leaf canopy. One is also able to distinguish between the grapes, stems, and leaves.The figure below shows (A) Visible image of three niagra grapes from June 18th. The toothpick in the picture is used to hold the sample fixed during the image acquisition. (B) Corresponding THz image. The image is 137 by 83 pixels. Each pixel is 0.3mm square. THz image is based on average reflection between 0.15-0.2THz of a grape (C) and grape hidden behind a grape leaf (D).
Ideally, one would like an accurate estimation of the crop yield months before harvest. Clearly based on the phase information, grape berries can be differentiated from stems and leaves. However, is this true throughout the growing season? To answer this question, we have measured the THz reflection images from grape clusters prior to bloom and after fruit set, stems and leaves from early May until late August. The figure below shows (A) Typical THz time-domain waveforms from bright reflection pixels of the stem (purple), leaf (gray), and flower (light blue) compared to the reference (dark blue). Inset: Visible image of a flower cluster on May 14 prior to bloom. (B) Comparison among the reference waveform (dark blue), the reflected THz waveform from a berry (June 1) in the inset (black curve), and the predicted THz reflection (purple) from the berry.Unlike the time-domain pulses of Fig. 3 from berries, the relative phase of the reflected pulses from the flowers, stems and leaves are essentially the same as the reference. In contrast, samples taken just two weeks later (figure B) behave quite differently with respect to the THz reflection. At this point in time, the grape cluster is post-flowering but pre-shatter. All of the flowers/berries are still present. During development, the moisture content should be different in some of the berries. Some berries will not mature and consequently at a later time will fall off the cluster (shatter). The time-domain waveforms of figure B show that the berries have begun to form. The figure shows the waveform from a berry of the inset in comparison to the reference reflection from a flat gold mirror. Clearly, the phase of the reflection from the berry is shifted by radians relative to the reference. The phase of the reflections from the stems and leaves, however, maintain the same relative phase as the reference. Based on the THz phase measurements, THz imaging can differentiate between berries and stems/ leaves very early in the growing season (June 1). In examining typical time-domain waveforms from grape clusters from June 1 through harvest, the general trend persists of a –π/2 phase shift from the berries relative to the reference. Time-domain waveforms from the stems/ leaves also show a consistent trend of roughly maintaining the same phase relationship as the reference waveform.
Since THz/ MMW wave imaging can be used to differentiate berries from the stems and leaves of a vineyard, a natural question is to what extent this is a generally applicable to estimation of yields of other crops. Our preliminary work on imaging of other common thin-skinned, high water content crops such cherries, blueberries, and plums show that the THz imaging method can be used to differentiate these fruits from the canopy as well. Moreover, by adjusting the focal distance from the THz lenses to the fruit in Eq. , it should be possible to design the THz/MMW optical components for a workable standoff distance. In the case of vineyards, a 1m distance is reasonable. By imaging high-water content fruits from through a thin canopy, THz/ MMW imaging may prove to be a useful method of crop estimation when either pruning techniques or canopy’s presence interferes with alternative crop estimation methods.