HYPERSPECTRAL IMAGING OF CONTAMINANTS IN PRODUCTS AND

PROCESSES OF AGRICULTURE

CROSS-REFERENCE TO RELATED APPLICATIONS

The present application claims priority from Australian Provisional Patent Application No 2005905552 filed on 7 October 2005, the content of which is incorporated herein by reference.

FIELD OF THE INVENTION

The present invention relates to an electromagnetic light method for characterizing a sample(s). In particular, the present invention relates to near infrared (NIR) and/or visible-near infrared (VIS-NIR) and/or UV imaging methods for characterizing sample(s).

BACKGROUND OF THE INVENTION Present methods for assessing quality indicators of natural and mass produced products often require extensive sample preparation and multi-step methodologies. In the field of agriculture, it is often desirable to analyse agricultural products, such as plants, plant parts, plant tissue and plant products, to determine one or more characteristics of interest. For example, agriculturalists involved in the production of fruit, use a number of both physical and chemical characteristics of the fruit to help forecast fruit quality, maturity and consumer preferences. Characteristics can include any one or more of the following ripeness, firmness, density, sugars, total soluble solids (TSS), pH, total acidity, density, anthocyanin, tannin, pigments including yellow, orange, brown and red, insect(s), infective agent(s), internal and external defects, foreign material.

Previous methods which have been used to assess fruit include for example, visual inspection and grading, tasting and laboratory analytical techniques. Visual methods of inspection tend to be time consuming and potentially subjective. Laboratory analytical techniques have major disadvantages resulting from the large amount of sample handling. The samples must be harvested, collected, bagged, labeled, dried, and finally sent to the laboratory, ground and analyzed for constituent analysis. This excessive sample handling adds both cost and time to the analysis. Agriculturalists would prefer to make informed decisions regarding the fruit before harvesting and these kinds of methods do not lend themselves to an assessment out in the field.

Viticulturists and winemakers are particularly interested in accessing analytical techniques which can quickly and efficiently assess grape qualities that can often be related to the characteristics of the wine produced. Furthermore, like many other agricultural products, grapes and grape vines are susceptible to a variety of infective agent(s) which can affect the grape vine, the grapes, and the wine produced from infected grapes. Infective agents may be microorganisms and may include viral, bacterial and fungal infection. Some infective agents may be pathogenic.

Of particular concern to viticulturists is fungal infection which can have major deleterious effects on both the plant and the agricultural products derived from the plants. For example, powdery mildew, caused by Erisyphe necator, is one of the most economically devastating diseases affecting grapevines (Vitis vinifera). Costly fungicide programs are widely used to prevent disease development on grapes. However, low levels of infection can persist on susceptible varieties in seasons with favourable weather conditions for disease, especially if sprays have been missed between flowering and berry set.

Powdery mildew not only reduces the yield and marketability of grapes, but also the quality of the wine. Levels of disease on grapes as low 3% can taint wines. Low levels of disease are difficult to assess visually in the vineyard and to quantify in large consignments of grapes. Another major disease of concern to Australian viticulturers is Botrytis bunch rot, caused by Botrytis cinerea. When conditions are favourable for disease, such as wet weather, grapes are often infected by Botrytis regardless of the application of spray programs and integrated disease and pest management strategies. Botrytis infection in harvested grapes can have a significant detrimental impact on wine quality through oxidative reactions caused by fungal laccase. The presence of fungal storage polysaccharides in juice and wine can also cause clarification problems. Often, secondary infections that cause other fruit rots and produce off-flavours in wine, are also associated with Botrytis rot.

Ochratoxin A (OA) is a mycotoxin that can cause kidney disease and affect the immune system of humans. It may also be carcinogenic and have teratogenic effects. In recent years, the presence of OA in food has been an issue of increasing importance and levels of tolerance for OA in a variety of foods are being established. Research has found that Aspergillus carbonarius, and sometimes Aspergillus niger (both black Aspergillus spp.) are present in Australia and may produce OA. The Aspergillus fungi involved in OA production appeared to be secondary invaders that infected grapes only

after damage caused by pre-harvest rain, infection by other fungi or mechanical damage.

Wineries worldwide are moving towards the assessment of grapes in the field prior to harvest and the assessment of grape loads at the weigh-bridge, as an aid for decision-making related to the handling of fruit for winemaking and to determine payments for grape quality. Grapes are rejected if the level of infection by infective agents is too high. However, these assessments can be time consuming, subjective and inaccurate, and some infestations can only be detected by the examination of berries with a microscope. This is especially true for the assessment of loads of machine- harvested grapes at the weigh-bridge. The subjective nature of these assessments can lead to disputes between the grape grower and the winery. Hence, the development and use of a rapid, quantitative and/or qualitative method for assessing levels of infection contamination in grapes in the vineyard and at the weigh-bridge would provide benefits for grape growers and winemakers. Another important assessment that is made at the weighbridge is the presence of

"matter other than grapes" (MOG) - this can include leaves, canes, vine-wood, trellising material, stones and is currently graded visually with the aid of reference charts. Hyperspectral imaging could make this more objective.

US Patent No. 6,847,447, the entire disclosure of which is incorporated herein by reference, describes apparatus and a method for measuring and correlating characteristics of whole fruit using the near infrared spectra. This approach is not only cumbersome and expensive; it is slow, as whole fruit is assessed one piece at a time and only after it has been harvested.

With current technology, samples must be collected, prepared then presented in a cuvette or sample cup to a NIR spectrometer. This is a cumbersome technique, not suited to weighbridge or field operation. Hyperspectral imaging offers the advantages of no sample preparation, the ability to analyse large sample areas/volumes and rapid real time analysis.

US Patent No. 5,991,025, the entire disclosure of which is incorporated herein by reference, describes an apparatus which combines an NIR spectrophotometer and a combine harvesting device. This arrangement provides real time analysis of the harvested agricultural product, in this case, corn and grain. However, analysis is only conducted on harvested products and the apparatus has limited applicability to the types of agricultural products and characteristics examined.

SUMMARY OF THE INVENTION

We have discovered that an NIR image and/or a VIS-NIR image can be used to determine the presence of an infective agent in a sample.

Accordingly, in a first aspect the present invention provides a method of determining the presence of "matter other than grapes" (MOG) associated with a sample(s), the method comprising: obtaining a near infrared (NIR) reflectance image(s) of at least a portion of the sample(s); and analysing the image(s) to determine the presence of "matter other than grapes" (MOG). Preferably, the MOG is the presence of one or more infective agent(s).

In a further aspect the present invention provides a method of determining the presence of one or more infective agent(s) associated with a sample(s), the method comprising: obtaining a near infrared (NIR) reflectance image(s) of at least a portion of the sample(s); and analysing the image(s) to determine the presence of one or more infective agent(s).

In another aspect, the present invention provides a method of determining the presence of one or more infective agent(s) associated with a sample(s), the method comprising: obtaining a visible-near infrared (VIS-NIR) reflectance image(s) of at least a portion of the sample(s); and analysing the image(s) to determine the presence of one or more infective agent(s). The sample may be an agricultural product. The agricultural product may be selected from one or more of the group comprising fruit, berry, bulb, grain, seed, leaf, flower, stem, vine, root, petal, and/or part thereof. Preferably the agricultural product is a fruit. Preferably the fruit is a red or a white grape.

The infective agent may be a microorganism. The microorganism may be selected from one or more of the group comprising a virus, bacteria, protozoa and/or fungus.

Preferably the microorganism is a fungus and may be selected from one or more the group comprising Erysiphe necator, Botrytis cinerea, Aspergillus fungi such as

Aspergillus carbonarius, Aspergillus niger; Ceratostomella ulmi; Claviceps purpurea; Xylaria mali; Xylaria polymorphs, Sclerotinia; Scleroderma; Tulostoma;

Synchytriaceae; Synchytrium endobioticum; Saprolegnia ferax; Pythium; Phytophthora citrophthora; Phytophthora infestans; Plasmodiophora brassicae; Clavariaceae; Hydnaceae; Basidiomycete, basidiomycetous; Lentinus edodes; Lentinus lepideus; Corticium salmonicolor; Corticium solani and /or other microorganisms in the fungus family.

The microorganism may be a pathogenic microorganism. In a preferred embodiment of the invention, the sample is a grape(s) and the fungal infection is by a fungus which causes powdery mildew in grapes. Preferably the fungus is Erysiphe necator. The determination may be quantitative or qualitative.

The method may include the use of a light source to illuminate the sample(s). The light source may be selected from any one or more of the group comprising tungsten halogen lamp, light emitting diode, laser diode, tuneable diode laser and / or flash lamp. According to another aspect, the present invention provides a method of determining one or more characteristics of an agricultural sample(s), the method comprising obtaining an NIR and/or VIS-NIR reflectance image of at least a portion of the agricultural sample(s); and analysing at least a portion of the image using chemometric analysis to determine one or more characteristics of the agricultural sample.

The characteristic(s) may be selected from one or more of the group comprising: sugars, total soluble solids, anthocyanin, tannin, pigments selected from the group comprising yellow, orange, brown and red; acidity; colour; pH; total acidity; firmness; internal and / or external disorder; the presence of infective agent(s); insect(s), and eating quality.

The method may be applied to the agricultural product prior to harvesting of the agricultural product. This application of the method may be used by the agriculturalist to assist in the management of the agricultural product, for example, fertilizer needs, whether or not to harvest the product, timing of harvesting, water needs and treatment with anti-infestation agents, such as fungicide agents.

The imaging method may be used during harvesting of the agricultural product.

The imaging camera may be mounted on the harvesting device and a real-time assessment of the characteristics of the product obtained. This kind of information is useful as the results could be used to determine whether the agricultural product meets

a predetermined standard. For example, the product may be separated at the time of harvest into different groups depending on the results of the imaging analysis.

The method may be applied during transportation of the agricultural product(s). The imaging equipment may be mounted onto the transport vehicle where one or more characteristics of the product(s) could be determined during transport of the product. The method may also be performed at the weigh-bridge where it could be used to assess the quality of the product and potentially used in determining the price paid for the product(s) and whether to accept the product(s). The method may be used to determine whether further processing of the product is required before transportation to the buyer of the product.

The method may be applied to the agricultural product during processing of the agricultural product. In the wine making industry, winemakers are desirous to assess one or more characteristics of the grapes before making the wine. For example, the level of infective agent(s) associated with the grapes, for example Erysiphe necator, can influence the quality of wine produced. Knowledge of the infective agent levels of the grape can be used to tailor the wine making process to suit the characteristics of the grape.

The method may be applied to detect the presence of foreign material with the product such as leaves, wood, stones, trellising material. The NIR reflectance image obtained comprises thousands of linearly independent, spatially-resolved NIR reflectance spectra which are collected with each collection. One or more of these individual spectra may be processed using chemometric analysis techniques to either qualitatively or quantitatively detect, classify, identify and/or visualize one or more characteristic(s) of the product(s). In yet another aspect, the present invention provides a method for sorting grape(s) according to the level of infection of an infective agent(s), the method comprising: obtaining a NIR and / or VIS-NIR reflectance image(s) of at least a portion of the grape(s); analysing the image(s) to assess the level of infective agent(s) present; comparing the level of assessed infection to a predetermined level and sorting the grape(s) accordingly.

With this embodiment, it is preferable that the infective agent is a fungus, most preferably the fungus is Erysiphe necator. The sorting may be performed before harvesting, during harvesting and/or after harvesting.

Analysis of the image according to the first and second aspects of the present invention may be performed using the analysis method of the third aspect.

BRIEF DESCRIPTION OF THE DRAWINGS

Figure 1 illustrates the plot of DNA content (ng/150 ng) vs visually graded infection level. Figure 2 (a) illustrates the plot of absorbance vs wavelength for raw spectra.

Figure 2 (b) illustrates the plot of the first derivative of absorbance vs wavelength.

Figure 3 (a) illustrates the plot of absorbance standard deviation for all samples vs wavelengths for raw spectra. Figure 3 (b) illustrates the plot of the first derivative absorbance standard deviation for all samples vs wavelengths.

Figure 4 (a) illustrates the plot of total soluble solids (°Brix) vs visually graded infection level.

Figure 4 (b) illustrates the plot of pH vs visually graded infection level. Figure 5 (a) illustrates the mean values of the first principal component (PCl) at each powdery mildew infection level, error bars show 1 standard deviation. Principal component analysis (PCA) was done on first derivative spectra with a wavelength range of 450 - 1884 nm.

Figure 5 (b) illustrates the mean values of the second principal component (PC2) at each powdery mildew infection level, error bars show 1 standard deviation. Principal component analysis (PCA) was done on first derivative spectra with a wavelength range of 450 - 1884 nm.

Figure 5 (c) illustrates the mean values of the third principal component (PC3) at each powdery mildew infection level, error bars show 1 standard deviation. Principal component analysis (PCA) was done on first derivative spectra with a wavelength range of 450 - 1884 nm.

Figure 5 (d) illustrates the mean values of the fourth principal component (PC4) at each powdery mildew infection level, error bars show 1 standard deviation. Principal component analysis (PCA) was done on first derivative spectra with a wavelength range of 450 - 1884 nm.

Figure 6 illustrates the correlation of NIR predicted and reference values for powdery mildew DNA content (ng/150 ng total DNA), using a PLS calibration with 1st derivative spectra using a wavelength range of 450 - 1900 nm: R ² = 0.98, standard

error of calibration (SEC) = 0.07 ng, standard error of cross validation (SECV) = 0.09 ng, 2 PLS factors used.

Figure 7 illustrates the prediction of Botrytis infection level of Shiraz grapes using PLS analysis of first derivative transformed grape spectra (600-1800 nm, R ² = 0.95, SECV= 1.25% infection).

Figure 8 illustrates the wavelength loadings for the first 3 factors of a PLS calibration for Botrytis infection of Shiraz grapes.

Figure 9 illustrates a plot of absorbance vs. wavelength for three grape wood E. lata culture extracts, with varying Eutypinol concentrations as measured by HPLC. Figure 10 illustrates a plot of PCl vs PC2 for principal component analysis of the UV spectra of unfractionated E. lata culture extracts. Samples are marked by eutypinol concentration (high= H, medium = M, low = L), as determined by HPLC.

DETAILED DESCRIPTION OF EMBODIMENTS OF THE INVENTION

The present invention will now be described with particular reference to the analysis of whole grapes for infective agent(s). However, it will be clear that a similar technique may be utilized for other samples, including manufactured goods, and for the analysis of other characteristics.

The present invention provides a system which is effective, fast, has high resolution, and which has a greater accuracy and discrimination rate than prior art devices or systems. It is desirable that the present invention provide a method and apparatus for the detection and discrimination of defects in products, such as agricultural products.

It is further desirable that the present invention provides a method and apparatus for sorting products based on the character, number, type or aggregation of defects.

The correlation of a disease state with structural and chemical changes can be established by challenging samples to be tested with the organism.

As used herein, an "agricultural product" refers to any plant material that is being interrogated by a method of the present invention. An agricultural product can be, for example, a fraction of a grape, a whole grape, more than one grape, and other plant tissues, among others. Controls can include grapes known to be susceptible and resistant. The correlation of the disease to a particular structural change can be established by an appropriate statistical analysis. It is understood that controls need not be run against a particular grape or grape batch once a correlation has been established.

Other agricultural products or plant tissues can be substituted for grapes. As used herein, agricultural products include, but are not limited to, plant tissues such as fruit, but also include non-plant based material such as non-organic matter or non-plant based matter that occur in an agricultural context. As used herein, plant tissues include, but are not limited to, any plant part such as leaf, flower, root, and petal.

NON-LIMITING EXAMPLES

Powdery Mildew Sample Collection

A set of 26 sample grape bunches, of approximately 200 g each, were collected from the 2003 vintage. The initial plan was to collect grapes at 5 infection levels (level 1 = uninfected, level 2 = 1-10%, level 3 = 11-30%, level 4 = 30-60% and level 5 = 61- 100% infected), with at least 4 replicates at each level. The samples were examined using an optical microscope and categorized, however, insufficient samples could be found at level 3 and 4 so they were combined to allow adequate replication. The samples were stored at -18°C until they were further processed.

Sample Preparation

Grape samples were thawed and then homogenised while still cool (approximately 4°C) with a Grindomix GM 200 (Retsch GmbH & Co., Haan, Germany) for 20 seconds at 8,000 rprn using a floating lid to maintain contact of sample with the blades. Homogenates were immediately sub-sampled and frozen for comparative DNA analysis. NIR scans were performed immediately on the remaining homogenates.

TSS and pH Analysis

To measure TSS (expressed as °Brix) homogenates were clarified by centrifugation then measured on a Palette digital refractometer (model PR 101, Atago

Co. Ltd. Japan). The pH of the grapes was measured on homogenates using an Orion

Advanced portable pH meter (model 250A, Thermo Orion, USA) equipped with an

Orion ROSS epoxy body combination electrode (model 815600, Thermo Orion, USA).

Quantification of Powdery Mildew in Grapes Using DNA Probes

The infection level of powdery mildew caused by the fungus Erysiphe necator, was assessed using a Erysiphe necator-spGciύc DNA probe, pUnAl as described by

Stammer, B.E. and Scott, E.S. (2001) "Detection of powdery mildew in grapes, must and juice." In: 'Abstracts of 11th Australian Wine Industry Technical Conference,

Adelaide SA'. (Poster No. P166) page 259.

DNA extractions were carried out using a CTAB extraction buffer, where 1-3 g of grapes (1-2 grapes) used for each replicate extraction. Grape samples were ground to a fine powder in liquid nitrogen before being suspended in CTAB buffer at 6O ⁰ C for 20 minute with gentle agitation. Following the addition of chloroform: isoamyl alcohol (24:1) and gentle agitation for 10 minute, the sample was separated into discrete phases by centrifugation. The DNA contained in the supernatant was then precipitated with ice-cold isopropanol before washing in 76% ethanol and subsequent storage in TE buffer at -2O ⁰ C.

Southern slot blot assays were prepared for all samples using the Bio-Rad Bio- dot SF unit as recommended by the manufacturer. Slots were hybridised with a

Erysiphe necator-specific probe, pUnAl, obtained from a plasmid library. Southern hybridisation, probing and autoradiography were according to Sambrook, J., Fritsch,

E.F. and Maniatis, T.A. (1989) 'Molecular cloning: a laboratory manual.' (Cold Spring

Harbor Laboratory Press, New York, USA). Slot blots contained approximately 50- 100 ng of total DNA (grapevine plus Erysiphe necator). Controls included DNA extracted from healthy grapevine tissue, and from various micro-organisms commonly associated with grapevines.

To quantify the Erysiphe necator DNA present in each sample, all slot blots also contained DNA (0.1-10 ng) extracted from conidia of Erysiphe necator produced on tissue-cultured plantlets. Slot Blot assay to detect DNA of Erysiphe necator and the probe clearly detected < 0.1 ng of powdery mildew DNA, which equates to < 1-5% infected bunch surface area or approximately 100 spores.

A strong relationship between powdery mildew DNA content and visual infection level is shown in Figure 1. Tukey pairwise comparison probabilities, from analysis of variance of DNA content and infection level, showed that all combinations of pairs of infection levels, other than level 1 and 2, had significantly different mean

DNA contents (p<0.001).

The entire DNA probe analysis process takes approximately 1 week to perform.

Vis-NIR Scanning and Data Analysis

Homogenates were scanned, without temperature equilibration, in a FOSS NIRSystems 6500 (FOSS NIRSystems, Silver Spring, Maryland, USA), in reflectance mode at 2 nm intervals over the wavelength range of 400-2500 nm. A reference scan was performed before each sample, using a rare earth metal oxide impregnated ceramic tile as a reference. Spectra were stored as the average of 32 scans. Scanning control

was performed with the Vision software package (FOSS NIRSystems, Silver Spring, Maryland, USA).

In order to make quantitative measurements or qualitative discriminations between samples, chemometric models were developed for each parameter. The model is a mathematical construct developed using samples of the same product or class of products. A Chemometric model is developed by collecting spectral readings from a group of samples that display (a) the maximum variability of the characteristic of interest, and (b) non-correlating or random variability in all other characteristics. The same samples are submitted for independent testing to measure the characteristic of interest by a standard analytical method. The spectral data and independent test data were then analyzed using commercially available chemometrics software. The statistical processes used in quantitative spectral analysis include multiple linear regression, classical least squares, inverse least squares, and principal component regression. The statistical processes used in qualitative spectral analysis include K- nearest neighbours, SIMCA and others.

When a sufficient number of samples were collected and properly analyzed, a mathematical model is constructed that describes the relationship between specific spectral features and the sample characteristic of interest.

Chemometric analyses were performed with WinISI software (version 1.5, Infrasoft International, LLC, USA). Discriminant analysis of principal component analysis (PCA) scores of spectra was performed with Systat (SPSS Inc., Chicago USA). Typically, spectra were pre-processed by smoothing with standard normal variate and detrend (4 data point gap) and transformed with the first derivative (4 data point gap). Reference data were either the visual classification level or the powdery mildew DNA concentration.

Reflectance spectra of uninfected and powdery mildew affected samples are shown in Figure 2. A total of 26 spectra are shown. Raw spectra Figure 2(a) show the dramatic baseline shifts typical of reflectance spectra, related to particle size effects.

Spectral features correlated with infection level can be observed both in the visible (400 - 700 nm) and NIR regions (700 - 2500 nm). The large variations due to baseline shifts with raw spectra are illustrated by the broad spectral standard deviation, Figure 3 (a). The spectral standard deviation profile is sharpened with first derivative transformation and some distinct areas of spectral variation occurred in both the visible and NIR wavelength ranges, Figure 3(b). An explanation for spectral variations with infection could be that TSS and pH can be correlated with infection level. This would apply particularly in the NIR

spectral regions. There appeared to be very variable but high TSS/low pH in the highest infection level Figure 4. Analysis of variance (ANOVA) with Tukey pair wise comparisons revealed that for TSS level 5 and 2 showed significant differences (Table l(a)). For pH, level 5 and level 1; level 5 and level 3 showed significant differences (Table l(b)). For the first 3 infection levels, there were no significant differences in TSS and pH; therefore, sugar or acid would not influence the NIR spectra when comparing levels.

Level 1 2 3 5

1 1.000

2 0.936 1.000

3 1.000 0.932 1.000

5 0.125 0.041 ^* 0.195 1.000

(a)

Level

1 1.000

2 0.660 1.000

3 1.000 0.773 1.000

5 0.007 ^** 0.075 0.016 ^* 1.000 (b)

Table 1 Tukey pair wise comparison probabilities for one-way analysis of variance (ANOVA), with infection level as the categorical variable compared with (a) total soluble solids (TSS) or (b) pH as the dependant variable.

Principal component analysis (PCA) was performed on first derivative spectra. Means and standard deviations for the first four principal components of spectra from each infection level are shown in Figure 5. The first principal component correlated strongly with the infection level and combinations of the other components provided further discrimination of infection levels.

A classification matrix for discriminant analysis of infection level, using the first four PCA scores is shown in Table 2(a); 100% classification was achieved. Note that a quadratic function was used (ie. the changes with infection level were not linear). If cross-validation was used (ie. samples sequentially removed from the training set and predicted independently), 92% classification was still achieved, suggesting that the classification algorithm is robust and not over-fitted to this relatively small dataset Table 2 (b).

Level 1 Level 2 Level 3 Level 5 % correct

(Predicted) (Predicted) (Predicted) (Predicted)

Level 1 7 0 0 0 100

Level 2 0 7 0 0 100

Level 3 0 0 5 ^■ 0 100

Level 5 0 0 0 6 100

Total 7 7 5 6 100

(a)

Level 1 Level 2 Level 3 Level 5 % correct

(Predicted) (Predicted) (Predicted) (Predicted)

Level 1 7 0 0 0 100

Level 2 0 6 1 0 86

Level 3 0 1 4 0 80

Level 5 0 0 0 6 100

Total 7 7 5 6 92

φ)

Table 2 Classification matrices for discrimination of powdery mildew infection level using the first 4 principal component analysis scores and a quadratic function for the best combination of scores to discriminate levels. 2 (a) shows data without cross- validation. 2 (b) shows data with cross-validation i.e. sequential removal of samples that were predicted with remaining samples. Predicted levels are in columns and actual levels in rows. Correct predictions are italicized. Only one level 2 and one level 3 sample (shaded in bold) was incorrectly classified during cross-validation.

Partial least squares analysis was performed on spectra to test an algorithm for prediction of the DNA content. A plot of NIR predicted and reference powdery mildew DNA content is shown in Figure 6. Using the 400 - 1900 nm wavelength range, the R ² was 0.98 and the standard error of cross validation was 0.09 ng DNA per 150 ng total DNA. This degree of accuracy is sufficient to clearly discriminate the lowest infection level (0-10%) from the uninfected samples. Similar calibration statistics were obtained if the wavelength range was restricted to 450 - 1098 nm: this wavelength range is achievable with relatively low cost silicon diode array-based instruments. Note that only two PLS factors were used as the calibration was very strong and not likely to be over-fitted.

Significant correlations between VIS-NIR spectra and powdery mildew infection level were observed, independent of basic grape composition. Infected grapes can be correctly classified on the basis of visual infection level using spectral information. Powdery mildew DNA concentrations correlated with visual infection levels and VIS-NIR calibrations can predict DNA levels. Botrytis

To overcome the problem of co-infection with other grape pathogens, as is often the case with in-field Botrytis infections, laboratory infected samples were prepared. Grapes were surface sterilized with hypochlorite, injected with a laboratory culture and incubated in humid trays at ambient temperature until sporulation was evident. Samples of varying infection levels (expressed as % infection w/w) were scanned over a 400- 2500 nm wavelength range. Calibrations were developed with PLS regression on first derivative transformed spectra. Botrytis has a dramatic effect on grape pigments, so to avoid domination of calibrations by visible wavelengths and thereby reducing calibration robustness in the face of grapes with varying intrinsic pigment concentrations, only NIR wavelengths were used (600-1800 nm). An example of a calibration for detection of Botrytis in Shiraz grapes had an R ² of 0.95 and a standard error of cross-validation of 1.25 % w/w. Calibration loadings were strongest at approximately 700 nm, a wavelength easily achieved with inexpensive silicon detectors.

In order to produce an electromagnetic light reflectance image of a sample(s), a Spectral Dimensions' NIR Chemical Imaging (NIR-CI) camera can be used. The camera can be tuned over the wavelengths of approximately 700 to 2500 nm and the imaging system utilizes an indium gallium arsenide (InGaAs) focal-plane array (FPA) detector comprising of 240 * 320 pixels for a total of 76,800 spectra per image cube recorded.

A camera which can be tuned over wide wavelengths can also be used. For example, a camera which can be tuned to include at least parts of the visible region can also be used. For example, the camera can be tuned to record VIS-NIR spectra from 350 to 2500 nm.

Non-limiting examples of suitable cameras include a Chemlmage CONDOR Macroscopic Chemical Imaging System camera and an Electrophysics Jade SWIR imaging camera.

The FPA can also be comprised of Si, SiGe, PtSi, InSb, HgCdTe, PdSi, Ge, analog vidicon types. The FPA output is digitized using an analog or digital frame grabber approach.

The electromagnetic light reflectance image of a sample(s) can be taken using ambient light from the sun as a light source.

Alternatively, the sample area can be illuminated using an appropriate artificial light source. Illumination with a light source can enable the rapid acquisition of reproducible data with good signal/noise (S/N), even in the highly light scattering and absorbing 250-699 nm and the strongly absorbing >950 nm region. The lamp can be a tungsten halogen lamp, for example a 12- Volt, 75-Watt tungsten halogen lamp. Other light sources which can be used include but are not limited to light emitting diode, laser diode, tunable diode laser, flash lamp and other such sources which will provide equivalent light source and will be familiar to a person skilled in the art.

The lamp is held at a resting voltage of 2- Volts. When a measurement is taken, the lamp is ramped up to the desired voltage, a brief delay allows the lamp output to stabilize, and then spectra can be acquired. After data acquisition, the lamp is ramped down to the resting voltage. This procedure extends lamp life and prevents burning the sample. In high speed operations the lamp can always be lighted, e.g., on a high-speed packing/sorting line or used on harvest equipment, and a light "chopper" or shutter or other equivalent article or method can be utilized to deliver light to the passing sample for a determined period of time. The operation of the light source is important in extending lamp life, reducing operating expense and reducing disruption of operations. The lamp voltage is ramped up and down to preserve lamp life and to lessen the likelihood of burning fruit. A standby voltage keeps the lamp filaments warm. An ambient/room light background measurement is made to correct for the dark spectrum, which can include ambient light. It is stored and subtracted from the sample and reference (if applicable) so that there is no contribution of ambient light to the sample spectrum, which would affect accuracy. Dual intensity illumination is employed to: 1) improve data accuracy above 925 nm and below 700 nm and 2) to normalize path length changes due to scattering. Dual exposure time increases the likelihood of increased data quality with large and small fruit. Utilization of more than one light detector, with each positioned at different distances from the sample, will likewise increase the ability to obtain increased data quality throughout each portion of the spectrum from approximately 350 nm to 1150 nm.

The electromagnetic light reflectance image comprises thousands of linearly independent, spatially-resolved spectra which can be collected with each collection. These spectra can then be processed to generate unique contrast intrinsic to analyte species without the use of stains, dyes, or contrast agents. For example, contrast can be generated in a VIS-NIR reflectance image and reveals the spatial distribution of

properties revealed in the underlying VIS-NIR spectra. Thus the acquired VIS-NIR reflectance image comprises many thousands of pixels with each pixel represents a full spectrum. This kind of data can be suitable for multivariate (chemometric) analysis techniques such as principal component analysis (PCA), principal component regression (PCR) and partial least squares (PLS) modeling, as discussed above.

Chemometric analysis of one or more of the spectra can then be used to determine one or more of the physical or chemical characteristics of the sample. In one application, chemometric analysis (or variants thereof such as piecewise direct standardization) are used to relate the spectra(s) to characteristic of the sample such as sugar composition and concentration, total soluble solids, anthocyanin, tannin pigments, including yellow and/or red coloured, acidity, pH, total acidity, firmness, color, presence of micro-organisms or foreign matter, internal or external disorder severity and type, and eating quality.

The acquired spectra(s) can be used to assess the level of infective agent(s) of the sample(s). For example, a VIS-NIR reflectance image can be analysed to assess the level of fungal infection of a fruiting body, such as a grape or berry.

Thus in order to assess the level of a infective agent, for example powdery mildew infection, an electromagnetic light reflectance image of the grape(s) can be taken and following analysis of the results, the level of powdery mildew of the grape(s) can either be qualitatively or quantitatively assessed. For example an NIR and / or a VIS-NIR reflectance image of the grape(s) can be used to assess the level of powdery mildew infection.

Other types of fungus which can be assessed using the present VIS-NIR imaging technique include any one or more chosen from the group comprising Ceratostomella ulmi (Dutch elm fungus - fungus causing Dutch elm disease); Claviceps purpurea, (fungus that infects various cereal plants forming compact black masses of branching filaments that replace many grains of the plant; source of medicinally important alkaloids and of lysergic acid); Xylaria mali (black root rot fungus, fungus causing black root rot in apples); Xylaria polymorpha (dead-man's-fmgers, the fruiting bodies of the fungi of the genus Xylaria); any fungus of the genus Sclerotinia (some causing brown rot diseases in plants); Scleroderma; Tulostoma; slime mold, slime mould - a naked mass of protoplasm having characteristics of both plants and animals; sometimes classified as protoctists; Synchytriaceae (aquatic fungus that causes pond scum); Synchytrium endobioticum (causes potato wart disease in potato tubers); Saprolegnia ferax, (attacks living fish and tadpoles and spawn causing white fungus disease); white rust (fungus causing a disease characterized by a white powdery mass of conidia);

Pythium; Phytophthora citrophthora (causes brown rot gummosis in citrus fruits); Phytophthora infestans (causes late blight in solanaceous plants especially tomatoes and potatoes); Plasmodiophora brassicae (resembles slime mold that causes swellings or distortions of the roots of cabbages and related plants); Clavariaceae (often brightly coloured that grow in often intricately branched clusters like coral); Hydnaceae (tooth fungus); Basidiomycete, basidiomycetous fungi; Lentinus edodes (Oriental black mushroom, shiitake, shiitake mushroom - edible east Asian mushroom having a golden or dark brown to blackish cap and an inedible stipe) Lentinus lepideus, scaly lentinus (a fungus with a scaly cap and white flesh and a ring on the stalk; Corticium salmonicolor (fungus causing pink disease in citrus and coffee and rubber trees etc); and Corticium solani (causes bottom rot in lettuce).

The assessment can be qualitative or quantitative.

The ultraviolet (UV) wavelength range

The visible wavelength range is useful for monitoring pigmented compounds hi natural products, for example grape anthocyanins. Near infrared can be used to monitor all compounds with hydrogen attached to carbon, nitrogen and oxygen, but in biological samples the spectrum tends to be dominated by a water signal. A wavelength range that provides further information, specificity and high sensitivity is the UV range. Wavelengths below 200 nm (the "vacuum UV" range) are difficult to use as absorbance is too high and all chemical bonds absorb, but the 200-400 nm range can be used to monitor compounds with double bonds such as tannins, proteins, anthocyanins, carotenoids, aldehydes, ketones, C-aromatic compounds, O and S- heteroaromatic compounds and N-heteroaromatic compounds. In the case of conjugated systems, the absorbance peaks are shifted to longer UV wavelengths and often as far as to visible wavelengths. The main advantage of using the UV range in grapes, for example, is that common constituents such as water, sugars and organic acids contribute very little to the spectrum, whereas minor constituents of interest such as tannins, anthocyanins and phenolic flavour compounds have a strong signal.

Some fungal metabolites also have a strong, unique UV fingerprint. Examples are the metabolites of Eutypa lata, a fungus that infects vine tissue resulting in the syndrome known as "dead arm" that causes serious economic loss in older vineyards. The acetylenic phenol metabolites produced by E. lata have an unusual structure with series of conjugated double bonds, triple bonds and aldehyde groups. This gives them a strong and unique UV spectral fingerprint and offers an opportunity for rapid

screening with minimal sample preparation using spectroscopy combined with chemometric methods.

Figure 9 illustrates examples of UV spectra of unfractionated culture extracts of E. Lata. Eutypinol concentrations were measured by HPLC and spectra from culture extracts showed clear features related to Eutypino ^' l, in particular the peaks at 262, 276 and 308 nm. Principal component analysis of first derivative transformed spectra shown in Figure 10, illustrates the discriminatory power of UV spectra with regard to cultures with low medium and high Eutypinol concentration as measured by HPLC.

UV imaging of samples requires a UV light source such as a xenon discharge, mercury or halogen lamp. The imaging camera must have sensors responsive to UV wavelengths. The UV spectral image can be combined with a visible image so that spatial data can be combined with chemical data generated from the UV spectra collected within each pixel of the UV image. UV, VIS and NIR images can be combined to cover a wavelength range of at least 200 to 2500 nm. Imaging algorithms can utilize all or part of that wavelength range.

An example of the use of UV imaging is the detection of E. lata in specimens of grapevine wood tissue and leaves. Samples are scanned then the spatial image overlaid with a false colour image generated by an algorithm to detect fungal mass or metabolites, or to detect degenerated vine tissue. This method can be both qualitative and quantitative. UV imaging can also be used to detect insects, animals and other foreign biological material in grape loads, using for example a protein UV signal. Grape secondary metabolites such as phenolic compounds can also be detected with UV imaging in grape loads, or on the vine with a portable device. Undesirable vine tissue in grape loads, such as leaves, petioles, canes and woody tissue can also be discriminated by their unique UV fingerprint.

There are several insects of concern which can affect the quality of the grape, and/or affect the vine. Non-limiting examples of insects which can be detected using aspects of the present invention include: Grape Berry Moth The grape berry" moth is a key pest of grapes that is distributed in the United

States east of the Rocky Mountains, and in eastern Canada.

The larvae of this insect can cause serious damage to commercial vineyards by feeding on the blossoms and berries. Damage by grape berry moth may increase mold, rots and numbers of fruit flies.

Webbing over blossoms and berries, and leaf flap cocoons are indicative of grape berry moth and can be visualized by a combination of a UV reflectance image and a visible image. Grape Phylloxera Grape phylloxera is native to eastern United States, but has been distributed to other grape regions of the U.S. and is also established in Europe where it is of great economic importance. The leaf galls caused by grape phylloxera are unsightly and do little damage, however, infestation of the roots can be difficult to control and can lead to decline of vines. Severe infestations can cause defoliation and reduce shoot growth. Hosts include cultivated and wild grapes.

The wingless forms of the insect are very small, yellow-brown, oval or pear- shaped, and aphid-like. The winged forms, which are less apt to be seen, are also aphid- like, except that wings are held flat over the back. The presence of grape phylloxera is best recognized by characteristic galls it produces on the leaves or roots. Leaf galls are wart-like, whereas, root galls are knot-like swellings on the rootlets, and may lead to decay of infested parts. The presence of leaf galls can be detected using methods according to the invention. Grape Rootworm

Larvae devour small roots and pit the surface of larger roots, causing an unthrifty condition of the plant, and reduction in yield. Vines may be killed in 3 or more years when damage is severe. Adults make chain-like feeding marks on leaves and may also feed on the surface of green grape berries. Hosts include wild and cultivated grapes.

The adult beetle is elongate oval, sub-cylindrical, dark reddish brown, clothed with short pubescence and is about 0.5 cm to 0.8 cm long. The larva is white, hairy, curved, with a brown head.

Chain-like leaf feeding damage by the adults is diagnostic and can alert growers to adult activity. Grape Flea Beetle Grape flea beetle is found in the eastern two-thirds of the United States. Adults eat buds and unfolding leaves, causing leaves to be ragged and tattered. Larvae feed on flower clusters and skeletonize leaves in a manner similar to adult rootworm feeding. Hosts include grape, plum, apple, quince, beech, elm & Virginia creeper.

Adults are dark metallic greenish-blue, jumping beetles about 0.5 long; larvae are brownish and marked with black spots; eggs are pale yellow, and fairly conspicuous on upper leaf surface or under loose cane bark.

Adults overwinter in protected areas around vineyards, and start feeding on interior of primary buds and opening grape leaves in early spring. Damaged buds will not develop into primary canes which can reduce yields. Grape Cane Girdler Grape cane girdler is common in central and eastern United States. Adults girdle canes with a row of punctures, that causes canes to break off at the girdled areas. It is only a minor pest on grape, preferring Virginia creeper. Hosts include grape and Virginia creeper.

The adult is a black snout beetle about 0.31 cm long. The grub is slightly larger when full grown, and is white with a brown head and legless. It is very similar in appearance to the closely related grape cane gall maker. Grape Cane Gallmaker

This insect produces noticeable red galls on new shoot growth just above nodes. Galls are usually found along vineyard borders near wooded trashy areas or at the ends of rows. If galls will be removed by pruning, it should be done by mid- July before emerging adults exit galls. Grape Root Borer

Grape root borer is potentially the most destructive insect attacking grapes in some parts of the world. Larvae of this insect tunnel into the larger roots and crown of vines below the soil surface. Borer damage results in reduced vine growth, smaller leaves, reduced berry size, and fewer bunches of grapes.

Adults are brown moths with thin yellow bands on the abdomen and resemble some paper wasps. The front wings are brown while hind wings are clear. Male moths fly about in a manner similar to wasps. Larvae are cylindrical, cream-colored, with three pairs of true legs near the head and five pairs of fleshy abdominal prolegs each bearing two bands of tiny hooks. Redbanded Leqfroller

Redbanded leafroller is an occasional pest of clusters and fruits, and its symptoms are very similar to grape berry moth. Larvae of this insect will feed on botli foliage and clusters. Unlike grape berry moth larvae, redbanded leafroller larvae do not crawl into the berry but remain concealed in webbing on the cluster stem and feed on the stem as well as berries.

The adult redbanded leafroller is a 1.25 cm long reddish-brown moth with small areas of silver, gold and orange. The moth is recognized by the red band extending across the front wings when at rest. The larva is a small, yellowish-green, unmarked caterpillar. The head capsule is the same color as the rest of the body.

Thus the method of the invention can be used to identify the presence of a number of insects in agricultural products. For example, the insects that can be identified can be selected from the group comprising caterpillar(s), grasshoppers, beetles, moth(s), moth pupa, Grape Berry Moth, Grape Phylloxera, Grape Rootworm, Grape Flea Beetle, Grape Cane Girdler, Grape Cane Gallmaker, Grape Root Borer, Redbanded Leafroller, scale insects, flies, fruit flies, aphids, midges or mealy bugs.

While the method of the invention has been described with particular reference to the analysis of agricultural products, it should be understood that the invention can be applied to the analysis of other sample types. For example, the method of the invention can be applied to the determination of the presence of microorganisms on other types of samples, such as hospital surfaces, kitchen surfaces, cooking surfaces, factory surfaces, and other surface where it is desirous to determine the presence of microorganism contamination. The method of the invention can be used in any field including veterinary, industrial or human medicine. In the case of human medicine, the present methods can be applied to determining the presence of pathogenic microorganism on for example surgical gloves.

Any discussion of documents, acts, materials, devices, articles or the like which has been included in the present specification is solely for the purpose of providing a context for the present invention. It is not to be taken as an admission that any or all of these matters form part of the prior art base or were common general knowledge in the field relevant to the present invention as it existed before the priority date of each claim of this application.

Throughout this specification the word "comprise", or variations such as "comprises" or "comprising", will be understood to imply the inclusion of a stated element, integer or step, or group of elements, integers or steps, but not the exclusion of any other element, integer or step, or group of elements, integers or steps.

It will be appreciated by persons skilled in the art that numerous variations and/or modifications may be made to the invention as shown in the specific embodiments without departing from the spirit or scope of the invention as broadly described. The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive.