WO2019201786A1

WO2019201786A1 - Optical Inspection and Sorting Machine, and Corresponding Method Thereof

Info

Publication number: WO2019201786A1
Application number: PCT/EP2019/059465
Authority: WO
Inventors: Benedict Mark Murray DEEFHOLTS
Original assignee: Bühler Uk Limited (Bukl)
Priority date: 2018-04-20
Filing date: 2019-04-12
Publication date: 2019-10-24
Also published as: GB202016718D0; GB2587511B; GB2587511A

Abstract

Proposed is an optical sorting machine (1) for sorting of granular objects (1.13) affected by a mold producing aflatoxin. Thus, the granular objects (1.31) can be at least partially contaminated (for example, hotspot contamination) by the highly toxic aflatoxin. The optical sorting machine (1) comprises feeders (1.11) for passing the objects (P) through a sorting zone (1.4). The sorting zone (1.4) comprises illumination means (1.6; 3, 3'), one or more optical detectors (1.7; 7) sensing reflected light or emitted fluorescent light (FE), and sorting means (1.5) for separating desired objects (1.131) showing no aflatoxin contamination to an accept stream (1.42) and undesired objects (1.132) showing an aflatoxin contamination to a reject stream (1.43).

Description

Optical Inspection and Sorting Machine, and Corresponding Method

Thereof

Field of the Invention

The present invention relates to the field of optical inspection and sorting machines for effecting relative separation between undesired objects and desired objects. The invention is particularly applicable to the sorting of a number of different agricultural products, e.g. maize, rice, peas and beans, etc., based on characterizing such products using fluorescence, in particular characterizing products having mycotoxin contamination, such as aflatoxin contamination, for example, in commodity products, including maize, grains, peanuts, coffee beans, tree nuts and other kernels. More particularly, the invention is applicable to the sorting of granular objects that are liable to be contaminated by a toxic mold that produces a mycotoxin, such as aflatoxin. The term "objects" is used herein in a wide sense as including all kinds of granular agricultural products such as maize, rice, peas and beans, etc., in the form of finer and coarser granular particles.

Background of the Invention

Aflatoxin is a mycotoxin, i.e. a toxic secondary metabolite produced by fungal mould that grows on the plan† in the field or due to poor handling of grain post- harvest. Mycotoxins are able to cause diseases and death in humans and animals by their toxic chemical compounds. There are over 300 known mycotoxins. Other mycotoxins that are commonly found in maize include deoxynivalenol (DON), fumonisin, zearalenone (ZEA), ocharatoxin and ergot alkaloids. All mycotoxins are a cause of concern because of their impact on human health. Aflatoxin is the most toxic. In nature, around 20 different aflatoxins can be identified. Aflatoxins are poisonous carcinogens that are produced by certain mold species of the genus Aspergillus (Aspergillus flavus, Aspergillus parasiticus, Aspergillus tamarii, Aspergillus nomius and other geni) that grow in soil, decomposing vegetation, hay, or grains. They can, for example, be found in improperly stored staple commodities such as maize, cassava, chili peppers, corn, cotton seed, millet, peanuts, rice, sesame seeds, sorghum, sunflower seeds, tree nuts, wheat, and a variety of spices. When contaminated food is processed, aflafoxins may enter the general food supply lines. They have been found in animal and human foods, especially also in feedstocks for agricultural animals. Animals fed contaminated food can pass aflafoxin transformation products into eggs, milk products, and meat. Aflafoxins are most commonly ingested. A distinction is made between more than 20 aflafoxins, e.g. Bi, B2, Gi, G2, or the derivatives Mi, M2, etc. The most toxic type of aflafoxin is Bi, which is also able†o permeate through the skin. Small quantities of aflafoxin are sufficient to make a batch unsafe†o eaf. In acute cases, if causes death. Long-term exposure has been found†o cause cancer, and exposure during early years has been shown to contribute†o irreversible stunting in children. Worldwide, if is estimated†o be responsible for up†o 155,000 cases of liver cancer a year. Globally, 160 million children suffer from stunting (International Agency for Research on Cancer (IARC), Press release 242, 201 6). There is a wide interest in sensitive methods and devices for the qualitative defection, quantitative measurement and selective separation of contaminated material in a variety of products. One of the characteristics for aflafoxins, in particular Bi and Gi, is their luminescence. Fluorescence and phosphorescence characteristics of aflafoxins Bi and Gi are known in the state of the art, in particular their fluorescence excitation and emission spectra,

phosphorescence spectra, phosphorescence decay times, and limit of detectability (see e.g. Van Duuren ef al., Luminescence characteristics of aflafoxins Bi, and Gi Anal. Chem. 40, 13, 2024-2027).

Specifically, for maize, there are five types of aflafoxin found - Bi, B2, Gi , G2 and Mi . As mentioned, of these, aflafoxin Bi is the most toxic. Jus† a few contaminated kernels can make an entire grain lot unsafe for human and animal consumption. The aflatoxin-producing mold Aspergillus grows best in ho† and humid conditions or where grain is stored in a high-mois†ure environment. Extreme weather conditions that cause plant stress make crops more susceptible†o fungal infections. These can also be a consequence of insect attack. As mentioned above, Aflafoxin can also be found in groundnuts, rice, oil seeds, sorghum nuts, chili peppers and dried fruit and can be present in milk from cattle that have been fed contaminated grains, but it is maize that is most vulnerable†o contamination. Over one billion tons of maize are grown annually. It is mostly used in animal feed, however people in sub-Saharan Africa, Southeast Asia and Latin America largely rely on it as food staple. Research into aflatoxin and its effects on health first goes back†o 1960, after 100,000 turkeys died in the UK. The cause was discovered to have been aflatoxin in the peanut meal the turkeys had been fed. In 2015, the World Health Organization (WHO) confirmed aflatoxin as the most dangerous chemical food safety hazard (WHO Estimate of the Global Burden of Foodborne Disease. 2015). Aflatoxin is known†o be a potent carcinogen and a leading cause of liver cancer, particularly in developing countries. Worldwide it is estimated to be responsible for up to 155,000 cases of liver cancer a year, mainly affecting the poorest - 87% of cases are in Africa, Southeast Asia and the Western Pacific. Acute exposure to high doses of aflatoxin in individual batches of maize can cause death. More common is long-term chronic exposure to non-lethal but harmful doses of aflatoxin. This affects whole communities and is a significant health problem in many developing countries. According†o a report by the International Agency for Research on Cancer (IARC), 500 million people in developing countries are at risk of chronic exposure to aflatoxins throughout their lifetimes. There is also a link between aflatoxin and stunting in children. Stunting is the impaired growth and development that children experience from poor nutrition and repeated infection. If affects both the physical and cognitive development of the child. The

abovemenfioned IARC study has established a connection between aflatoxin poisoning and stunting. Children in countries that rely on a maize-based diet are regularly exposed to the poison either through the food they eat themselves or, as infants, through their mother’s milk. If can be passed to humans through milk from cattle that have been fed on contaminated maize, and it has been found in livers and eggs of chickens that have eaten contaminated grains. This problem occurs particularly in India, Pakistan and Asia. Aflatoxin also impairs the health and productivity of farm animals. For instance, young poultry that consume high levels of aflatoxin Bi in their feed may die due to liver damage or show poor growth performance, requiring more feed to gain weigh†. Aflatoxin also reduces the reproductivity of hens. They take longer to reach maturity and lay fewer and smaller eggs. Far more detrimental is aflatoxin Bi contamination in dairy animals like cattle, buffalo, goats, sheep and camels. Up to 3% is carried over†o the milk in the form of aflatoxin Mi, putting milk consumers at risk.

Aflatoxin itself is a chemical compound that is colorless and odorless. A consumer cannot detect its presence in food. To safeguard health, most countries in the world have rulings setting maximum permitted quantities for mycotoxins, and specifically for aflatoxins, in food and animal feed. However, since it is technically so difficult†o defect, high levels of contamination can go unnoticed. For example, a study by the International Food Policy Research Institute showed that the aflatoxin

contamination in maize in Kenya in 2010 was alarming - 30 to 40% of maize collected from farmers had levels greater than 10 ppb. In the case of maize stored by the farmers, 60% of it was above the limit of 10 ppb. Likewise, testing of 600 maize samples from principle retail markets in Rwanda showed means between 8 and 25 ppb. Aflatoxin is the only mycotoxin for which there are legal limits for animal feed as well as food.

Today, there is no solution to eliminate aflatoxin contamination from maize. Methods exist to prevent the fungal infestation by molds of the genus Aspergillus, but once the crop is affected, there is no treatment that destroys it. Instead, contaminated grains must be painstakingly removed from every batch. As few as 2 highly

contaminated grains in 10,000 healthy ones can render an entire batch unfit for purpose, and the process of identifying and removing the highly poisonous grains is still rather imprecise. For example, to be on the safe side, today’s maize cleaning removes tons of good grain, contributing to the loss of one third of food globally, while maize is the world’s main cereal crop and one third of the global population relies on it as a staple. Aflatoxin contamination of food is recognized by the Food and Agriculture Organization, the World Health Organization (WHO) and other public health bodies as one of the major technical challenges for food safety and security. Thus, there is a huge demand for effective inspection and sorting machines, such as optical sorting machines, in particular measuring-data sensitive systems, for example for crops such as maize that improves on current cleaning practices and has the potential to make a significant contribution to addressing the aflatoxin challenge.

Mycotoxins are stable chemical compounds that cannot be destroyed through cooking or other thermal processing. The mitigation strategy is two-pronged, focusing both on preventing fungal growth as well as eliminating contaminated grain. As it is not possible to eradicate the toxin entirely in any single step, a management system that tackles the problem along the whole supply chain is thus the best practice. The first step is to prevent the toxin formation through good and monitored agricultural practices in the field such as careful seed selection, crop rotation, protection from pests and proper application of fungicides. An emerging method is biocontrol, which involves spraying strains of the Aspergillus mold that do not produce aflatoxin on the field to create competition for the toxic strains. To reduce the risk of contamination at the grain handling and storage stage, it is important to ensure grain is dried to a safe moisture content. Failure†o dry grain correctly poses the biggest risk and is one of the main reasons for contamination as the mold thrives in moist environments. In areas where posfharvesf technology is no† available, this is a problem. Silos must be well ventilated†o prevent moisture condensation. Pes† control also plays an important role as pests make small holes in the grain where the mold can easily begin to grow. For feed materials, the costly application of organic acids is sometimes used†o prevent fungal growth and formation of aflatoxin. At the milling state, grain is cleaned mechanically†o remove fractions that are contaminated. The current state of the art relies on indirect indicators of fungal infection to guide which grains should be removed. Such indirect indicators for maize are shown in figure 5, wherein figure 5a shows contaminated dust, figure 5b contaminated black spotted maize, figure 5c contaminated green mold maize, figure 5d contaminated small broken maize, figure 5e contaminated dark brown maize, figure 5f contaminated shriveled brown maize and figure 5g contaminated grain with insect bites often having internal infections and/or low-density grains. When maize is attacked by fungal infection, the grains can become discolored, shriveled or shrunken. Contaminated grains break more easily and the dust within a contaminated batch of maize normally has a high level of contamination. Flowever, some grains show no external signs of infection and, depending on the batch of maize, the indicators may vary, with some indicators being caused by completely different fungi that cause no toxin. In the state of the art, typically, a combination of mechanical separation and optical sorting is used †o reduce levels of contamination. Firs†, a classifier sieves out broken kernels (cf. fig. 5d, 5a) . Then the maize passes through an aspirator to remove as much dust as possible (cf. fig. 5a) . I† then passes through a gravity separator to remove lighter and shriveled kernels (cf. fig. 5f), before finally passing through an optical sorter to remove discolored and drier kernels using a combination of visible and infrared cameras for detection (cf. fig. 5b, 5c, 5e). I† is shown that each of these methods can contribute†o the overall reduction in mycotoxin contamination. Up†o now, this combination of preventing, reducing and constantly monitoring and measuring has been the only effective approach available. Flowever, even in doing so, i† does no† eradicate aflatoxin altogether. For this reason, there is an ongoing need†o find better and/or more efficient solutions for reducing mycotoxins in food and feed chains. The state of the art today for aflatoxin contamination reduction is by a combination mechanical cleaning, using an aspirator to remove dust, a classifier†o remove broken grains and a gravity separator†o remove shriveled lightweight grains. Another problem is that the products can only be sorted if it has been assessed, typically through sample testing in a laboratory, as potentially having an unacceptable level of contamination. Such state of the art approaches achieve contamination level reductions of between 60% and 90% at the cost of removing between 5-25% of product from the food chain. Since such prior art systems rely on sample faking and batch testing, they remain incapable of providing reliable actual indicators of contamination; in particular, they are not able to provide real-time monitoring. Another technical problem (hotspot) of aflatoxin contamination within a bulk of product is that hotspots make it difficult to achieve accuracy, especially by sample methods such as batch testing. A solution†o this lies in providing for continuous monitoring of all products in order†o reduce risks from sampling errors. Various inspection apparatus and methods have been suggested in the prior art for detecting aflatoxin, but it had remained impossible to efficiently sort aflatoxin- contaminated product. An example is shown in Hesseltine et al (1973), "New methods for rapid detection of aflatoxin", Pure Appl. Chem., 35, 259-266,

10.1351 /pad 97335030259, disclosing the application of UV in detecting aflatoxin and the importance of detecting the correct color of fluorescence. Further, GB-A-993063 discloses the use of UV fluorescence to detect aflatoxin, and highlights the problem associated with weak emission signals from the UV excitation. US-A-4535248 discloses the use of long-wave UV excitation for detection of aflatoxin in blanched almonds. US A-4866283 discloses the use of laser excitation at one wavelength and detecting the level of induced luminescence at a second wavelength†o identify aflatoxin in roasted peanuts. US-A-8841570 discloses the use of a wideband UV light source and a RGB camera for detecting the presence of almonds and detecting aflatoxin by looking at the fluorescence level in the red channel of the RGB camera. US-A-8563934 discloses the use of UV LEDs or fluorescent lamps as a light source and detecting peak fluorescence and peak fluorescence shit† features within the emitted spectra in order†o identify aflatoxin contamination. EP-A-2270475 discloses the use of a UV lamp, preferably a Wood's lamp,†o provide illumination and an RGB camera to view the product, with the number of pixels considered to be contaminated being counted and referenced to a calibration curve in order†o provide a measure of aflatoxin content. Other prior art includes CN-A-103234945, CN-A-104646315, CN-A-105044062, US-A- 3305089, US-A-2010/0193412, US-A-2012/0093985 and US-A-2012/0280146. Up†o now, with no consistent visible difference between the kernels that are contaminated and those that are not, direct analysis with optical systems has proven imprecise, inefficient, and energy-consuming. As mentioned, the current standard method of testing for aflafoxin and measuring contamination levels is to take a sample and send it to a laboratory†o be ground down, extracted with a solvent and analyzed by wet chemistry. There are several related problems with this. One of the major challenges is that just a few contaminated kernels can make an entire grain lot unsafe— as few as 2 contaminated grains in 10,000— and the distribution of these is not homogenous but occurs in hotspots. This means, for example, that in a truck loaded with grain, samples taken from different parts of the load and analyzed will reveal widely divergent values. As one sample alone cannot provide sufficient information about the average level of contamination in the overall load, it is necessary†o take many samples. However, sampling fakes time, as each sample must be milled, extracted and sen††o a laboratory. These procedures are a bottleneck in aflatoxin analytics. In practice, it means that results arrive too late to reliably support real-time decisions in processing. An alternative is to measure the extracted sample with a rapid test that requires very basic laboratory infrastructure. However, even if food processors rely on rapid analysis of several samples, there is still a good chance of missing contamination hotspots. Given the difficulties in accurately identifying where aflatoxin contamination is present in maize, food processors err on the side of caution, throwing out large quantities of clean grain. Even so, occasional contaminated batches still slip through†o the next stage in processing. If the grains are ground, it becomes impossible †o remove contamination from parts of a lot. If contamination is found at this stage, the whole lot has to be thrown away and the losses mount. Since these techniques are broadly aimed at detecting and removing indicators of fungal infection, rather than targeting the toxin directly, the results are variable. Typically, this process achieves a reduction in toxin levels of between 60 and 90% with total yield losses varying between around 5 and 25%. Another problem is the rising probability of occurrence, as climate change brings higher temperature and drought†o more regions of the globe. In the past, aflatoxin contamination was mainly found in tropical regions. Today, it is appearing in regions where it was previously uncommon or unknown. A 2°C temperature increase due to climate change is predicted to significantly increase aflatoxin also in colder areas, such as Europe. Aflatoxin is already a significant risk†o a large par† of the world’s population. With climate change, the number of people exposed to that risk is growing. Thus, there is an enormous demand for technical solutions enabling fast, certain and efficient food processors with better decision making in real-time.

One major improvement is the introduction of fluorescence as a presumptive test for aflatoxin in maize. In particular, the green fluorescence associated with aflatoxin shows a very bright green color. Figures 6a/6B/6C exemplarily show images with contaminated grain (fluorescent green) and healthy grain (fluorescent blue). A and B are the same image a† different exposures. C shows 3 contaminated grains mixed with healthy grains. In the present inventive system, there is a new way of providing a new kind of sorting system using the color of the fluorescence. If is observable that clean grains fluoresced a blue color under ultraviolet light and that all kernels that fluoresced a greenish color showed high levels of aflatoxin when tested.

The difference lies in the fact that grains contaminated by aflatoxin also contain a substance called kojic acid, which is also produced by the Aspergillus mold. If is kojic acid that is responsible for the bright green color. With this much more specific indicator of aflatoxin contamination, the challenge is†o provide a sorting machine that can accurately defect and identify the precise difference in color between contaminated and healthy grains a† industrial speeds, and which can additionally be operated efficiently.

Summary of the Invention

If is an object of the present invention to provide an optical inspection and sorting machine and appropriate method therefor, no† having the disadvantages of the prior art. In particular, with the new sorting machine, losses from sorting should only occur if aflatoxin is present. More particularly, if granular objects such as maize are sorted, no residual loss should arise from false positives. The new systems shall make it possible†o apply the possible sorting over the whole sorting process, without having the disadvantage of high energy consumption and inefficiency. The sorting machine shall make it possible†o apply a direct analysis with optical systems without having the disadvantage of being imprecise or of taking samples and sending them to a laboratory to be ground down, extracted with a solvent and analyzed by we† chemistry. The sorting machine should be efficiently applicable†o batches; the distribution of (few) contaminated kernels is no† homogenous but occurs in hotspots. In particular, in such inhomogeneous cases, where typically one sample alone cannot provide sufficient information about the average level of contamination in the overall load, and overall sampling would take unproportional time, the invention should be applicable efficiently, both in use of time and energy, thereby keeping control over the whole batch.

According to the present invention, these objects are in particular achieved by the features of the independent claims. In addition, further advantageous embodiments can be derived from the dependent claims and related descriptions. According to the present invention, the above-mentioned objects related

†o the optical sorting machine are achieved, particularly, in that for sorting of granular objects, the optical sorting machine comprises a hopper and/or feeders for passing the objects†o a sorting zone, the sorting zone comprising illumination means, one or more optical sensors sensing reflected or emitted light (FE) from the objects, and sorting means for separating desired objects†o an accept stream and undesired objects†o a reject stream, in that the optical sorting machine comprises optical splitter means for measuring luminous intensities of the reflected and/or emitted light, or a photometric equivalent thereof sensed by the optical sensors, in at leas† a firs† and a second wavelength range, in that the optical sorting machine comprises trigger means for triggering trigger identifications of undesired objects based on their luminous intensities measured in the a† leas† firs† and the second wavelength range of the emitted light, by means of a plurality of defined sets of a† leas† 2-dimensional intensity trigger vectors, wherein each defined set of intensity trigger vectors defines an intensity color space of trigger identifications of objects by luminous intensities in the a† leas† firs† and second wavelength ranges, wherein at leas† one of the intensity color spaces comprises trigger identifications of objects with a higher probability of correctly identifying undesired objects than at leas† one other of the intensity color spaces, in that the optical sorting machine comprises a frequency counter for measuring the frequency or number of undesired objects identified in the a† leas† one of the intensity color spaces with the higher probability of triggering undesired objects, and comprises measuring means for measuring an overall frequency or number of objects passing the sorting zone, and in that the optical sorting machine comprises a steering device for controlling operation of the sorting means by means of two switched modes switching between sorting and monitoring operation of the sorting machine, wherein if a predefined threshold value of the frequency or number of undesired objects identified in the at least one of the intensity color spaces with the higher probability as a portion of the measured overall frequency or number of objects passing the sorting zone is exceeded, then the sorting means are switched on†o sorting mode, while otherwise the sorting means are switched off to only monitoring mode of the optical sorting machine.

As an embodiment variant, the triggered undesired objects of the at least two intensity spaces can be objects contaminated by aflafoxin producing the fluorescence excitation, wherein the granular objects are affected by a mold producing aflafoxin. The optical sorting machine can for example comprise an interface module connecting the optical sorting machine to the world-wide backbone network for automated accessing of data related to the overall risk level or overall probability for aflafoxin contamination in the geographic area that the optical sorting machine is located in and/or that the granular objects to be sorted come from, wherein the measured probability for triggering undesired objects is improved. The first wavelength range can for example be between 450 - 495 nm in the blue light range and the second wavelength range is between 495 - 570 nm in the green light range. The one or more optical detectors can for example at least comprise cameras for splitting the received light info two wavelength ranges using a prism with a dichroic filter element. The means for causing the objects to fluoresce can for example comprise at least one incident light emitter emitting one or more incident light beams causing the fluorescence of the objects. The emitted incident light can for example comprise ultraviolet (UV) light. The one or more detectors sensing the emitted fluorescent light can for example comprise optical filters for isolating incident light and fluorescent light. The optical filters for isolating incident light and fluorescent light can for example comprise filter fluorometers using filters to isolate the incident light and fluorescent light or spectro-fluorometers using diffraction grating monochromators to isolate the incident light and fluorescent light. The one or more optical detectors can a† least for example comprise cameras. The sorting means can for example comprise one or more ejectors ejecting undesired objects from the objects to be sorted, giving the accept stream and the reject stream. The one or more ejectors at least can for example comprise pneumatic ejectors. The invention has inter alia the advantage that the sorting machine will only have losses from sorting if aflatoxin is present, in particular if the objects are sorted there will be no residual loss from false positives. With the new system, when the risk is low the system is able to switch automatically from sorting†o monitoring. When sorting†o remove aflatoxin using the sorting machine, there are some contaminated kernels that are easily identified with a high contrast relative to the good kernels, and there is another group where the difference is much smaller and if this group is removed, a higher yield loss is provided. The present invention makes it possible to use the high- confrast kernels as indicators and turn the sorting on and off depending on these to reduce the yield loss. In other words, the present invention allows for using two different color or identify spaces that have high contrast relative to the good material as indicators of aflatoxin. The number of objects containing these color spaces is counted together with the number of kernels of maize or granular objects that have passed through the sorter. A measurable indication of the number of contaminated grains as a proportion of the good grains is generated, and thus a measurable indication of whether the product is contaminated can be derived. As a variant, the inventive system can also be enabled by combining this with a connection†o the cloud to access and download relevant data on the overall risk level for aflatoxin in the area where the sorter is located or where the product came from. Combining this data will improve the measured risk. In general, the present invention provides a new sorting technology minimizing toxic contamination in maize and improving yield, by identifying and removing cancer-causing, aflatoxin-infected grains. The inventive apparatus does this more accurately and at greater speed than any previous prior art solutions. The inventive sorting machine provides a significant advance for the maize processing industry in its fight against fungal molds called mycotoxins, the most poisonous of which is aflatoxin. The inventive sorting machine is able to eliminate up to 90% of

contaminated maize. The invention provides advances in digital technology, together with advances in sorting and food safety, which makes the inventive system contribute †o solving a major global food safety and security challenge. Until now, sorting granular objects, e.g. maize, for aflatoxin reduction has proved difficult and imprecise, relying on identifying indirect indications of contamination. Testing for contamination based on sampling is inconclusive and time-consuming, as contamination occurs in hotspots. Just two contaminated kernels in 10,000 are sufficient†o make a lot unfit for purpose.

Alongside health risks, the economic impact on farmers and food processors is significant. The inventive sorting machine is the first optical sorting technology able to identify aflafoxin based on direct indicators of contamination, while simultaneously using real-time, cloud-based data to monitor and analyze contamination risk. The invention operates based on triggering the color each kernel fluoresces as it passes under powerful UV lighting in the sorter. If is known that contaminated kernels fluoresce a specific bright green color. The inventive machine’s highly sensitive cameras detect this precise color of fluorescence. Within milliseconds of detection, air nozzles deploy to blow contaminated kernels out of the product stream. The machine processes up to 15 tons of product an hour, eliminating up to 90% of contamination - a significant improvement on current solutions. As an embodiment variant, a cloud-based grid structure is a key enabler†o reducing overall yield loss. Combining data from the cameras with data stored in the cloud allows a local, real-time analysis of the risk of contamination†o be carried out. When the risk is minimal, sorting is halted while the machine continues to monitor. If the risk rises, sorting automatically restarts. The inventive machine, coupled with the cloud data access, reduces yield loss to below 5%, compared with between 5% and up to 25% for other solutions.

Brief Description of the Drawings

The present invention will be explained in more detail by way of example in reference to the drawings in which: Figure 1 schematically illustrates an embodiment of an optical sorting machine 1 for sorting of granular objects (P) that comprises feeders 1 .2 for passing the objects (P) through a sorting zone 1 .4, the sorting zone comprising illumination means 3, 3', one or more optical sensors 1 .7, 7 sensing reflected 1 .62 or emitted 1 .63 light (FE) from the objects (P), and sorting means 1 .5 for separating desired objects to an accept stream and undesired objects to a reject stream.

Figure 2 schematically illustrates an embodiment of an inspection apparatus 2 of the sorting zone 1 .4 of the optical inspection and sorting machine 1 for inspecting and sorting a flow of product/objects 1 .41 (P), the inspection apparatus 2 being suitable for use with the present invention. Figure 3 illustrates a vertical sectional view through the inspection apparatus 2 of the sorting machine 1 of Figure 2.

Figure 4 is a schematic diagram depicting, as an example, relevant intensify color spaces 31 , 32, 33, 34 of maize illustrating the functionality of the present invention. The x-axis gives the intensify of the measured blue color wavelength range, while the y- axis gives the intensify of the measured green color wavelength range. Reference numerals 32 and 33 depict intensify color spaces where the triggered objects show a bright defect, i.e. having a high probability of being undesired objects 1 .132. Reference numeral 31 depicts an intensify color space where the triggered objects show a good product trigger identification, i.e. having a high probability of being desired objects 1 .131 . Reference numeral 34 depicts an intensify color space where the triggered objects show similar characteristics of being desired objects 1 .131 and undesired objects 1 .131 , i.e. in the defect product color trigger space, the trigger identification of the undesired objects 1 .132 is similar to trigger identifications of good product (desired objects 1 .131 ) overlapping with good product color trigger space 31 . Thus, figure 4 illustrating the exemplary intensify color spaces for maize on a sorter 1 , there are two color spaces 32/33 where the bright defects appear, a good product color space 31 and a defect space 34 that overlaps the good color space 31 . When defects in the overlapping defect color space are targeted, good material, i.e. desired objects 1 .131 , is also ejected. If the product contains aflafoxin, there will be some defects in the bright defect color spaces 32/33, and the measured trigger identifications are used as distinct indicators of contamination by the optical sorting machine 1 .

Figures 5 exemplarily shows indirect indications of aflafoxin contamination in maize no† based on luminosity measurements as used by prior art sorting machines. Figure 5.1 shows contaminated dust, figure 5.2 contaminated black spotted maize, figure 5.3 contaminated green mold maize, figure 5.4 contaminated small broken maize, figure 5.5 contaminated dark brown maize, figure 5.6 contaminated shriveled brown maize and figure 5.7 contaminated grain with insect bites often having internal infections and/or low-density grains. The current state of the art relies on such indirect indicators of fungal infection (as shown in figures 5)†o detect which grains should be removed. When maize is attacked by fungal infection, the grains can become discolored, shriveled or shrunken. Contaminated grains break more easily and the dust within a contaminated batch of maize normally has a high level of contamination. However, some grains show no external signs of infection and, depending on the batch of maize, the indicators may vary, with some indicators being caused by completely different fungi that cause no toxin. In the prior art, a combination of mechanical separation and optical sorting is typically used†o reduce levels of contamination. Firs†, a classifier sieves out broken kernels. Then the maize passes through an aspirator to remove as much dust as possible. I† then passes through a gravity separator to remove lighter and shriveled kernels, before finally passing through an optical sorter to remove discolored and drier kernels using a combination of visible and infrared cameras for detection. Figures 6 exemplarily shows three images with contaminated grain

(fluorescent green) and healthy grain (fluorescent blue) . A and B are the same image at different exposures. C shows 3 contaminated grains mixed with healthy grains.

Detailed Description of the Preferred Embodiments Figure 1 shows an optical sorting machine 1 (also called optical color sorting machine), wherein the sorting is based on the measured characteristics differences of granular materials, using an optical sensor 1 .7†o drive sorting means 1 .5†o sort different granular materials 1 .41 (P) . The sorter 1 can comprise a hopper 1 .1 with granular material 1 .41 to be sorted. Further, the sorter 1 can comprise one or more feeders 1 .1 1 , in particular vibratory feeders 1 .1 1 1 feeding the material 1 .41 from the hopper 1 .1 1 to the one or more chutes 1 .12, 1 .121 /1 .122/... /1 12x providing the sorting zone 1 .4 with the object stream 1 .41 . Other feeders 1 .1 1 , for example bowl feeders or sorters with belt- type transport processes, are also imaginable for the inventive system. It is†o be mentioned that the present optical sorter 1 can be used widely in many industry applications, such as agricultural products and the food industry. The most common use is sorting granular objects 1 .41 , e.g. maize, rice, wheat, corn, peanuts, a variety of beans, seeds, tea, herbs, dried vegetables, etc. The sorting machine 1 is operated according to the difference in the optical properties of the material 1 .41 using photoelectric detection technology by getting and triggering red and/or green and/or blue color information, in particular identity measurements of the corresponding wavelength ranges and/or depth identification of small and fine impurities. The sorting machine 1 sends steering signals to the sorting means or ejector 1 .5 controlling its operation. The machine 1 can for example use compressed air-driven ejectors

(pneumatic ejectors 1 .51 ) for automatic sorting of the desired objects 1 .131 from undesired objects 1 .132. The identified undesired objects 1 .132 can for example be objects contaminated by aflatoxin, wherein the objects 1 .132 are affected by a mold producing aflatoxin.

The sorting zone 1 .4 comprises illumination means 1 .6; 3, 3', wherein the illumination means 1 .6 cause the objects 1 .132 to fluoresce and can for example comprise a† leas† one incident light emitter emitting one or more incident light beams. The optical detectors 1 .7 are configured†o detect fluorescence emitted from the objects 1 .132. The incident light beam(s) can tor example comprise ultraviolet (UV) light. The sorting machine 1 , sensing the fluorescence emitted by the optical detectors 1 .7, can for example comprise optical filters 1 .81 for isolating the incident light beam(s) and the emitted fluorescence. The optical filters 1 .81 for isolating the incident light beam(s) and the emitted fluorescence can for example comprise filter fluorometers using filters †o isolate the incident light and fluorescence or spectrofluorometers using diffraction grating monochromators†o isolate the incident light and fluorescence. In particular, the optical sorting machine 1 comprises the optical splitter means 1 .81 for measuring luminous intensities of the reflected 1 .62 and/or emitted 1 .63 light, or a photometric equivalent thereof sensed by the optical sensors/detectors 1 .7; 7, in a† leas† a firs† and a second wavelength range 1 .631 , 1 .632. 1 .63x. The firs† wavelength range 1 .631 can for example be between 450 - 495 nm in the blue light range and the second wavelength range (1 .632) can for example be between 495 - 570 nm in the green light range. The one or more optical sensors/detectors 1 .7 a† leas† comprise cameras 1 .71 for splitting the received reflected or emitted light into two wavelength ranges using a prism with a dichroic filter element.

The one or more optical detectors 1 .7 can a† leas† comprise cameras 1 .71 . The sensed fluorescence is the light given off or emitted by the undesired

contaminated objects 1 .132 when they are illuminated by ultraviolet light (UV) . Figure 6 shows images with contaminated grain (fluorescent green) and healthy grain

(fluorescent blue). A and B are the same image a† different exposures. C shows 3 contaminated grains mixed with healthy grains. I† is a very fain† light that is normally observed in a darkened room. One problem is that the color can only be defected by the camera using a long exposure. Under normal sorting conditions, where the cameras scan at a rate of 10,000 images a second, it is not possible to detect the color. A solution for the present invention is the use of a hyperspectral camera to gather large spectral data cubes of maize samples under UV light. A hyperspectral camera is able to break up color into up to 200 wavelengths, whereas a conventional camera works only with red, green and blue. Hyperspectral cameras capture around 1 gigabyte per data cube. This is the same amount of data in each line as a conventional camera captures for a whole picture. To achieve the spectral data processing for the steering device 1.9, object 1 .3 samples, for example maize, can be analyzed for example using a toximet analyzer or the like. A toximet analyzer is a highly accurate analysis that is more reliable than a strip test but does not require a full wet chemistry test in a laboratory. The standard toximet procedure is typically designed for 20 mg samples. In the present case, this procedure can be revised to account for single kernel measurement in order to determine contamination levels. The spectral data can for example be analyzed to find a correlation between the specific fluorescence color of the kernels and the

contamination levels. This data can then be used to design the spectral components for a camera to specifically target the green color associated with the fluorescence of the kojic acid that indicates the presence of aflatoxin.

Since the light levels associated with the fluorescence are normally at a very low level and cannot or can hardly be detected with the standard lighting system, a powerful LED-based ultraviolet lighting system 1.64 as illumination means 1 .6 can for example be used, along with a highly sensitive camera 1.71. However, combined with appropriate optical sensor/detector systems 1.7, lasers 1.65 or lamps 1 .66, in particular xenon arcs 1.661 or mercury-vapor lamps 1 .662, can also for example be used as illumination means 1 .6, if an appropriate level of fluorescence detection and measurement is possible. The lighting and camera system 1.6/1.7, as described above, is incorporated into the inventive optical sorting machine 1. After adjustments and modifications, in tests the inventive system achieved a stable reduction in aflatoxin contamination levels averaging 85-90% with a yield loss of less than 5%. These results are based on using the inventive optical sorting machine 1 alone, without adding other mechanical steps described above. However, in practice all steps can be used, thus improving the reduction in aflatoxin contamination levels and yield loss further.

Depending on the embodiment variant, the optical sorting machine 1 can be extended by additional the types of sensors, so that the optical sorter 1 is able†o recognize other properties of the objects 1 .13†o be sorted, such as color, size, shape, structural properties and chemical composition. Such sensors can for example comprise color cameras with high color resolution capable of defecting millions of colors†o better distinguish subtle color defects or trichromatic color cameras (three- channel cameras) dividing light into three bands, which can include red, green and/or blue within the visible spectrum as well as IR and UV. Coupled with image recognition, the optical sorting machine 1 that features such cameras can provide additional recognition of each object's color, size and shape as well as the color, size, shape and even location of a defect on an object, making it possible†o improve the elimination rate of contamination of the objects and reducing the yield loss further. The sorter 1 can further be configured†o additionally compare the objects 1 .13 to user-defined accepf/rejecf criteria to identify and remove defective products or foreign material from the sorting zone 1 .4 based on these other criteria, or to separate products of different grades or types of materials. The optical sorting machine 1 requires a compatible combination of light sources, i.e. the illumination means 1 .6†o illuminate objects and the optical sensors 1 .7†o capture images of the objects 1 .31 before the images can be processed, for example capturing the luminous intensifies in the a† leas† firs† and the second wavelength range 1 .631 , 1 .632 of the emitted light 1 .63, and the accept or reject decision made is made by the steering device 1 .9.

The optical sorting machine 1 comprises trigger means 1 .2 for triggering trigger identifications 1 .221 of undesired objects 1 .132 based on their luminous intensities measured in the a† leas† firs† and the second wavelength range 1 .631 , 1 .632 of the emitted light 1 .63, by means of a plurality of defined sets of a† leas† 2-dimensional intensity trigger vectors 1 .21 . Each defined set of intensity trigger vectors 1 .21 defines an intensity color space 31.34 (see figure 4) of trigger identifications 1 .22 of objects by luminous intensities in the a† leas† firs† and second wavelength ranges 1 .631 /1 .632. At leas† one of the intensity color spaces 32, 33 comprises trigger identifications 1 .22 of objects with a higher probability of correctly identifying undesired objects 1 .132 than a† leas† one other of the intensity color spaces 31 ,34. The intensity color spaces 31.34 are typically characteristic for a certain product†o be processed by the sorting machine 1 . Figure 4 shows relevant intensify color spaces 31 , 32, 33, 34 a† the example of maize as granular objects 1 .13; P†o be sorted. The x-axis gives the intensify of the measured blue color wavelength range, while the y-axis gives the intensify of the measured green color wavelength range. Reference numerals 32 and 33 depict intensify color spaces where the triggered objects show a bright defect, i.e. having a high probability of being undesired objects 1 .132. Reference numeral 31 depicts an intensify color space where the triggered objects show a good product trigger identification, i.e. having a high probability of being desired objects 1 .131 . Reference numeral 34 depicts an intensify color space where the triggered objects show similar characteristics of being desired objects 1 .131 and undesired objects 1 .131 , i.e. in the defect product color trigger space the trigger identification of the undesired objects 1 .132 is similar to trigger identifications of good product (desired objects 1 .131 ) overlapping with good product color trigger space 31 . Thus, figure 4 illustrating the exemplary intensify color spaces for maize on a sorter 1 , there are two color spaces 32/33 where the bright defects appear, a good product color space 31 and a defect space 34 that overlaps the good color space 31 . When defects in the overlapping defect color space are targeted, good material, i.e. desired objects 1 .131 , is also ejected. If the product contains aflafoxin, there will be some defects in the bright defect color spaces 32/33 and the measured trigger identifications are used as distinct indicators of contamination by the optical sorting machine 1 .

The optical sorting machine 1 further comprises a frequency counter 1 .31 for measuring a frequency or number of undesired objects 1 .132 identified in the a† leas† one of the intensity color spaces 32, 33 with the higher probability of triggering an undesired object 1 .132, and comprises measuring means 1 .32 for measuring an overall frequency or number of objects passing the sorting zone 1 .4. The optical sorting machine 1 comprises a steering device 1 .9 for controlling the operation of the sorting means 1 .5 by means of two switched modes switching between sorting and monitoring the operation of the sorting machine Uf a predefined threshold value of the frequency or number of undesired objects 1 .132 identified in the a† leas† one of the intensity color spaces 32, 33 with the higher probability as a portion of the measured overall frequency or number of objects passing the sorting zone 1 .4 is exceeded, then the sorting means 1 .5 are switched on to sorting mode 1 .931 , while otherwise the sorting means 1 .5 are switched off†o only monitoring mode 1 .932 of the optical sorting machine 1 . In an embodiment variant, the optical sorting machine 1 comprises an interface module 1.91 connecting the optical sorting machine 1 to the world-wide backbone network (Internet) 1.92 for automated access to location-specific data

1 .9212, 1.922.1 .922x related to the overall risk level or overall probability for aflatoxin contamination in the specific geographic area 1.94 in which the optical sorting machine 1 is located and/or that which the granular objects 1.13 to be sorted come from and/or that in which at least one other sorting machine 1 is located, whereby either intensity color spaces 31.34 for triggering the identification of objects are modified based on the accessed data 1.921 1 , 1.921.1.92x1 or the predefined threshold value is modified based on the accessed data. As a variant, the optical sorting machine 1 comprising the interface module 1.91 and connecting the optical sorting machine to the world-wide backbone network (Internet) 1.92 can for example create an alert over the Internet connection 1.92†o warn an operator of the sorting machine 1 that the product contamination may be too high, resulting in a higher probability that the accept product 1.42 is contaminated, too, where the frequency of occurrence of objects 1.132 in the at least one intensity color space 32, 33 is higher than a second threshold. In addition, the sorting machine 1 can for example report details of its mode of operation, either sorting or monitoring,†o the cloud-based database, and this information can for example be used for technical billing

procedures of a customer, further providing a report†o the customer, or both. Further, data on sorting modes uploaded to a cloud-based database 1.921 by a plurality of sorting machines 1 can for example be used to assess the generated risk of

contamination in the local area 1.94 and/or be used for creating alerts to other users in the geographical area 1.94. The novel cloud-based structure and cloud interference discussed herein can for example be realized using cloud-infrastructure for example provided by Microsoft or others.

As an additional embodiment variant, data in the databases 1.921 , 1.922...

1.92x can for example comprise localized measuring data such as weather condition parameters for that year from weather measuring stations, operational and sensory data from other sorting machines 1 in the same geographical area 1.94 and/or measuring data received from sensors in the silos where the objects 1.41†o be sorted (e.g. maize) 1.41 are stored. These data can be processed to provide parameters of merit x^merit from the processed data. These parameters are accessible to the optical sorting machine 1. The optical sorting machine 1 uses these parameters of merit x^merit as a multiplier for the threshold used†o automatically defect or trigger if product is contaminated or no†. The table below shows an example assuming that the original threshold is, for example, 0.1%, i.e. if more than 0.1% of the grains are in the obvious color class, then the sorting is automatically turned on. If the optical sorting machine 1 is connected†o the network 1.92 and is receiving probability data, then this threshold can for example be modified as set out in the table below, thus if the probability of contamination is high, the threshold can for example be reduced†o 0.05%, and if the probability of contamination is low, then the threshold can for example be increased†o 0.2%.

(Table 1 : Threshold modification and dynamic adaption)

The present inventive optical sorting machine makes it possible†o capture and analyze data relating to actual contamination and risk in the most efficient way. The large data sets captured by hyperspectral imaging of infected grains, combined with analytics procedures and laboratory back testing, made it possible†o provide the technical solution needed for automated sorting of the infected grains from the good ones. Connecting the machine†o a cloud-based infrastructure allows data†o be collected from the optical sorting machine 1 and an intelligent, device-driven assessment of risk to be taken, so that the optical sorting machine 1 only monitors the grain when the actual measured risk of contamination is low, i.e. below a predefined threshold value. As the object stream 1.41 flows through the sorting zone 1.4 of the optical sorting machine 1 , the transmitted electronic signals from the cameras 1.71 are monitored and the signal data on each object 1.3 is collected. This data can be combined with overall risk data transferred†o the optical sorting machine 1 and the steering device 1.9, respectively, allowing it to assess the overall level of risk of contamination, generated as a probabilistic parameter value, which can be used†o trigger on, initiate and/or activate and/or conduct automated decision making and steering operation. If there is a measured risk exceeding the trigger threshold, the optical sorting machine 1 sorts the object stream 1.41 , for example the flow of maize within the sorting zone 1 .4. If the risk becomes particularly high, the operator can for example automatically be alerted, allowing him†o take additional precautions. If the object stream 1 .41 , for example the feed maize, is assessed by the steering device 1 .9 to be clean and the overall risk is measured to be small, the optical sorting machine 1 can decide it is safe to signal halt sorting†o the sorting means 1 .5. As soon as the optical sorting machine 1 identifies aflatoxin by measurement in the object stream 1 .41 , i.e. the fed product, again or a higher level of risk, sorting is automatically resumed by appropriate signal generation and transfer†o the sorting means 1 .5. The intelligent automated decision-making can for example happen on a‘per-grain’ basis in less than 100 ms, at industrial grade throughputs of 10-15 tons per hour. Such a complete surveillance and intelligent sorting process cannot be achieved up to now by any known prior art sorting or cleaning system with the efficiency and optimized yield loss as provided by the present inventive optical sorting machine 1 . With the inventive cloud- based optical sorting machine 1 , in particular with the cloud-based optical sorting machine 1 , constantly monitoring the object flow 1 .41 (P) and constantly measuring and assessing risk, operators can now safely place sorters 1 in their process lines in the knowledge that yield losses should be almost zero, except for those rare occasions when the product is contaminated. If the object stream 1 .41 is measured as containing aflatoxin, the sorting procedure will be automatedly enabled, and the safety of the final accept stream 1 .42 should be assured.

The inventive optical sorting machine 1 provides a significant improvement on previous prior-art systems, enabling precision processing that reduces the risk of aflatoxin contamination while reducing food waste and operational risk. Thus, the invention provides the technical basis to have a widespread acceptance throughout the food and/or maize value chain. Grain handlers and elevators, faking in maize from the farmer, need to dry if and clean if preferably before storage. Using the optical sorting machine 1 according†o the present invention results in more consistent cleaning performance and a reduction in the losses associated with aflatoxin reduction. Bulk handlers of crop or maize at ship loading/unloading can use the optical sorting machine 1 as a high-capacity sampling system. In existing prior art systems, typically, a 10 kg sample is taken for analysis. The inventive optical sorting machine 1 is able to monitor a sample at 10-15 fons/hour in a flow of 200 tons/hr. This greatly reduces sampling and measurement errors and improves the security and safety of the product being shipped or unloaded from ships. Further, food and feed processor systems can also increase the safety of crops or maize with the inventive optical sorting machine 1 , achieving a 90% aflafoxin reduction through targeted defection and elimination of the crop or maize kernels that carry high contamination. This represents a significant reduction in risk. At the same time, losses are minimized†o below 5%. Today's prior art food and feed processor system generally rely on their supply chain to provide clean product rather than implementing conventional cleaning lines. This is largely due†o the associated yield losses. However, since testing comes with a high margin of error, due to the inhomogeneous distribution of aflafoxin, such prior art systems run the risk of getting lots that exceed legal requirements. With conventional sorters there will always be a yield loss whether the product is contaminated or no†. With the inventive optical sorting machine 1 , and the ability†o continuously measure and assess the risk of contamination, these systems are able†o monitor constantly or most of the time, and only start removing grains if contamination is detected, ensuring the safety of the final product. A 90% reduction of aflafoxin in contaminated maize lots means a significant reduction of risk. For example, it means reducing the contamination level from 20 ppb †o 2 ppb, which makes maize fit for human consumption. Feed raw material with an aflafoxin contamination level of 30 ppb can by sorted†o the level that is acceptable for dairy cattle (< 5 ppb)†o ensure healthy animals and safe milk. Overall, the operational risk for operators, for example farmers and grain handlers, is contained by having a better estimation of the level of aflatoxin contamination and by being able†o reduce it through precision sorting. It prevents losses due†o downgrading of maize†o feed or biomass quality. Beyond its potential†o reduce the impacts of aflatoxin, which mainly affect the food and feed industry in Europe and North America, the inventive optical sorting machine 1 can make a major contribution to reducing the health impacts in less affluent communities in Africa, India and South East Asia.

Figure 2 shows another embodiment variant of an optical inspection and sorting machine 1 comprising an inspection apparatus 2 for inspecting a flow of product 1 .41 (P), and which provides for characterization of product P, in this embodiment identification of product P having mycotoxin contamination, such as aflatoxin contamination. The inspection and sorting machine 1 also comprises ejectors 1.5, which are actuated†o effect ejection of undesired objects 1.132. In this embodiment the product P is maize grains, but could be other commodities, such as peanuts, coffee beans, tree nuts or other kernels, etc. The inspection apparatus 2 comprises a light source 3, which provides an elongate band of illumination I of a firs† wavelength or range of wavelengths, a camera 7, which has a viewing axis X that intersects the flow of product P, for receiving fluorescence emission FE a† a second wavelength or range of wavelengths, different from the firs† wavelength or range of wavelengths, along the viewing axis X from objects within the flow of product P, a dichroic mirror 9 that receives the band of illumination I from the light source 3 and provides the illumination I to the flow of product P along the viewing axis X of the camera 7 and receives fluorescence emission FE from objects within the flow of product P and provides the fluorescence emission FE †o the camera 7 along the viewing axis X of the camera 7, and an optical lens 15 which receives the illumination I from the light source 3 and concentrates the illumination I as a narrow, concentrated line of light to the dichroic mirror 9.

The light source 3 comprises a plurality of light source elements 3' that provide the band of illumination I. The light source elements 3' are arranged in an elongate line, here in adjacent relation on a printed circuit board. With this

arrangement, the position of the light source elements 3' is fixed and so no

maintenance of the plurality of the light source 3 is required. The light source elements 3' comprise light-emitting diodes (LEDs). In another embodiment, the light source elements 3' could comprise laser diodes (LDs). With this configuration, by arranging for the illumination I to be along the viewing axis X of the camera 7, a concentrated line of light can advantageously be employed, without concern for a depth of objects within the flow of product P, as illumination of the product P is ensured, as compared†o a configuration employing off-axis illumination. Illumination I is in the UV. Illumination in the UV has, for example, application in relation to the detection of mycotoxin

contamination, in that mycotoxin indicators fluoresce under UV illumination. The illumination I has a wavelength of less than 400 nm, here about 365 nm, and the dichroic mirror 9 reflects light at a wavelength of less than 400 nm and transmits light a† a wavelength which is greater than 400 nm. In an alternative embodiment, the illumination I could be in the IR or visible. For example, illumination in the visible red wavelength, here from 620†o 740 nm, has application in relation to the detection of chlorophyll, in that chlorophyll fluoresces under illumination in the visible red

wavelength, with the fluorescence emission FE being in the IR. The dichroic mirror 9 is configured such that the illumination I and the fluorescence emission FE are in a common viewing plane VP along the viewing axis X of the camera 7, here within ± 5 degrees of the viewing plane VP. The dichroic mirror 9 also acts as a cut filter for the illumination I, thereby cleaning up the illumination I. The dichroic mirror 9 acts as a highpass or lowpass filter, but no† as a bandpass filter, as employed in US-A-8563934. In this regard, if is imporfanf†o recognize that the

fluorescence emission FE from objects within the flow of product P is no† reflected or transmitted light, but rather light caused by fluorescence of the product P engendered by excitation with the illumination I, and a significant feature of the present application of the dichroic mirror 9 is that the fluorescence emission FE, in having a wavelength or range of wavelengths which is different from the wavelength or range of wavelengths of the illumination I, is transmitted to the camera 7, whereas the illumination I, such as could be transmitted or reflected by objects within the flow of product P, is reflected and prevented from passing to the camera 7. The camera 7 is a line-scan camera and is a multichromatic camera that is capable of viewing at least two selected

wavelengths for detecting the spectral response or signature of the fluorescence emission FE. The camera 7 may be a bichromatic camera that is capable of viewing two selected wavelengths. In an alternative embodiment, the camera 7 could be a hyperspectral camera, which allows for detection of the spectral response or signature of the fluorescence emission FE. One such hyperspectral camera is disclosed in

WO 2002/048687. The optical lens 15 is a rod lens.

The inspection apparatus 2 is housed in a light box 19. The inspection apparatus 2 further comprises a data processing device 21 for receiving signals from the camera 7 and processing the signals in characterizing the product P, in this embodiment for mycotoxin contamination. The processor 21 characterizes the product P for mycotoxin contamination by analysis of the spectral signature of the fluorescence emission FE. The characterization can be a likelihood of contamination. In one embodiment, the processor 21 is a computer analysis system, typically based on a PC.

In another embodiment, the processor 21 can be performed by hardwired logic, such as logic gate arrays. Thus, the inventive sorting machine 1 can be one using cameras that split received light into two wavelengths using a prism with a dichroic filter element.

Figure 4 schematically depicts color spaces that are used in analyzing each granular object 1 .41 (P) that passes through the inspection and sorting machine 1. When sorting is active, eject/non-eject signaling is taken with respect to the color spaces depicted. The bottom left-hand corner of the graph of figure 4 represents an intensity that was dark in both the green and the blue, and the top right-hand corner is saturated in both green and blue. Non-con†amina†ed kernels generally have intensities tiffing within a larger fairly central color space indicated by a first ovoid 31 , as illustrated in the diagram. Kernels which are contaminated on the outside tend to have little blue content and their intensity generally falls within a color space nearest to the left-hand axis in the diagram, indicated by a second ovoid 32. Broken kernels that are

contaminated are almost saturated in the green and their color space is depicted by a third ovoid 33 nearest to the top axis. Some grains are contaminated on the inside, and these tend to have a similar color†o some of the good kernels; their intensities tend to lie in a smaller central color space indicated by a fourth ovoid 34.

Finally, it will be understood that the present invention has been described in its preferred embodiments and can be modified in many different ways without departing from the scope of the invention as defined by the appended claims. By way of example, although described primarily in its preferred embodiments as having application†o the detection of mycotoxin contamination, and especially aflatoxin contamination, the present invention could be employed in other applications, such as to characterize chlorophyll content, for example, in seeds to determine viability, ochratoxin in coffee, or zearalenone in various cereal grains.

References

Optical inspection and sorting machine

1 .1 Hopper with granular material to be sorted

1 .1 1 Feeders

1 .1 1 1 Vibratory feeders

1 .1 12 Bowl feeder

1 .12 Chu†e(s)

1 .121 Chute 1

1 .122 Chute 2

1 .123 Chute 3

1 .12x Chute x

1 .13 Granular objects (P)†o be sorted

1 .131 Desired object

1 .132 Undesired object (contaminated)

1 .2 Trigger means

1 .21 Set of a† leas† 2-dimensional intensity trigger vectors

1 .22 Triggers identification of objects

1 .21 1 Trigger identifications of undesired objects

1 .212 Trigger identifications of desired objects

1 .3 Counter and frequency measuring means

1 .31 Frequency counter

1 .32 Measuring means

1 .4 Sorting zone

1 .41 Unsorted object stream (P)

1 .42 Accept stream

1 .43 Reject stream

1 .5 Sorting means / Ejector

1 .51 Pneumatic ejector

1 .51 1 Compressed air

1 .6 Illumination means

1 .61 Light beam (UV, ultraviolet (100 nm - 400 nm))

1 .62 Reflected light 1.63 Emitted light (FE, fluorescence)

1.631 First wavelength range

1.632 Second wavelength range

1.63x x-th wavelength range

1.64 LED (Light-Emitting Diode)

1.65 Laser (Light Amplification by Stimulated Emission Radiation)

1.66 Lamps

1.661 Xenon arcs

1.662 Mercury-vapor lamps

1.7 Optical sensor/detector

1.71 Camera

1.71 1 Flyperspectral camera

1.72 Light detector

1.721 Single-channeled light detectors (monochromatic) 1.722 Multi-channeled light detectors (multichromatic)

1.8 Signal processor and ejector controller

1.81 Optical splitter means (Incident light and fluorescent light splitter/isolator)

1.81 1 Filter fluorometers

1.812 Spectrofluorometers

1.9 Steering device

1.91 Interface

1.92 World-wide backbone network

1.921 Decentralized database 1

1.921 1 Localized aflatoxin occurrence pattern data

1.922 Decentralized database 2

1.9221 Localized aflatoxin occurrence pattern data 1.92x Decentralized database x

1.92x1 Localized aflatoxin occurrence pattern data 1.93 Switched modes

1.931 Sorting and monitoring operation

1.932 Monitoring only operation

1.94 Geographic areas

1.941 Geographic area 1

1.942 Geographic area 2 1 .94x Geographic area x

2 Inspection apparatus

21 Data processor

3 Light source

3' Light source elements

7 Camera

9 Dichroic mirror

15 Optical lens

19 Light box

3x Intensity color space comprising trigger identifications of objects 1 .13 31 Good product color trigger space of desired objects 1 .131

32/33 Bright defect color trigger spaces of undesired objects 1 .132 34 Defect product color trigger space similar or overlapping with good product color trigger space 31

Claims

1 . An optical sorting machine (1 ) for sorting of granular objects (1 .13; P) that comprises a hopper (1 .1 ) and/or feeders (1 .1 1 ) for passing the objects (1 .13; P) to a sorting zone (1 .4), the sorting zone (1 .4) comprising illumination means (1 .6; 3, 3'), one or more optical sensors (1 .7; 7) sensing reflected (1 .62) or emitted (1 .63) light (FE) from the objects (1 .13; P), and sorting means (1 .5) for separating desired objects (1 .131 )†o an accept stream (1 .42) and undesired objects (1 .132) to a reject stream (1 .43), characterized in that the optical sorting machine ( 1 ) comprises optical splitter means (1 .81 ) for measuring luminous intensifies of the reflected (1 .62) and/or emitted light

(1 .63), or a photometric equivalent thereof sensed by the optical sensors (1 .7; 7), in at leas† a firs† and a second wavelength range (1 .631 , 1 .632. 1 .63x), in that the optical sorting machine (1 ) comprises trigger means (1 .2) for triggering trigger identifications (1 .221 ) of undesired objects (1 .132) based on their luminous intensities measured in the a† leas† firs† and the second wavelength range (1 .631 , 1 .632) of the emitted light (1 .63), by means of a plurality of defined sets of a† leas† 2-dimensional intensity trigger vectors (1 .21 ), wherein each defined set of intensity trigger vectors (1 .21 ) defines an intensity color space (31.34) of trigger identifications

(1 .22) of objects by luminous intensities in the a† leas† firs† and second wavelength ranges (1 .631 /1 .632), wherein at leas† one of the intensity color spaces (32, 33) comprises trigger identifications (1 .22) of objects with a higher probability of correctly identifying undesired objects (1 .132) than at leas† one other of the intensity color spaces (31 ,34), in that the optical sorting machine (1 ) comprises a frequency counter (1 .31 ) for measuring a frequency or number of undesired objects (1 .132) identified in the a† leas† one of the intensity color spaces (32, 33) with the higher probability of triggering undesired objects (1 .132), and comprises measuring means (1 .32) for measuring an overall frequency or number of objects passing the sorting zone (1 .4), and in that the optical sorting machine (1 ) comprises a steering device (1 .9) for controlling the operation of the sorting means (1 .5) by means of two switched modes switching between sorting and monitoring the operation of the sorting machine (1 ), wherein if a predefined threshold value of the frequency or number of undesired objects (1 .132) identified in the at least one of the intensity color spaces (32, 33) with the higher probability as a portion of the measured overall frequency or number of objects passing the sorting zone (1.4) is exceeded, then the sorting means (1 .5) are switched on †o sorting mode (1.931 ), while otherwise the sorting means (1.5) are switched off to only monitoring mode ( 1.932) of the optical sorting machine ( 1 ) .

2. The optical sorting machine (1 ) according†o claim 1 , characterized in that the identified undesired objects (1.132) are objects contaminated by aflafoxin, wherein the objects (1.132) are affected by a mold producing aflatoxin.

3. The optical sorting machine (1 ) according†o one of claims 1 or 2, characterized in that the optical sorting machine (1 ) comprises an interface module

(1 .91 ) connecting the optical sorting machine (1 ) to the world-wide backbone network (1 .92; Internet) for automated access to location-specific data (1 .9212, 1 .922.1 .922x) related to the overall risk level or overall probability for aflatoxin contamination in the specific geographic area (1.94) in which the optical sorting machine (1 ) is located and/or that which the granular objects (1 .13) to be sorted come from and/or that in which at least one other sorting machine (1 ) is located, whereby either intensity color spaces (31.34) for triggering the identification of objects are modified based on the accessed data (1.921 1 , 1.921.1.92x1 ) or the predefined threshold value is modified based on the accessed data.

4. The optical sorting machine (1 ) according†o one of claims 1 to 3, characterized in that the optical sorting machine (1 ) comprises an interface module (1 .91 ) connecting the optical sorting machine to the world-wide backbone network (1 .92; Internet), where when the frequency of occurrence of objects (1 .132) in the at least one intensity color space (32, 33) is higher than a second threshold, the sorting machine will create an alert over the Internet connection (1.92)†o warn an operator of the sorting machine (1 ) that the product contamination may be too high, resulting in a higher probability that the accept product (1.42) is also contaminated.

5. The optical sorting machine (1 ) according†o one of claims 1 to 4, characterized in that the optical sorting machine (1 ) comprises an interface module (1 .91 ) connecting the optical sorting machine (1 ) to the world-wide backbone network (Internet), wherein the sorting machine (1 ) reports details of its mode of operation, either sorting or monitoring, to the cloud-based database, and this information being used for a technical billing procedure of a customer, providing a report to the customer, or both.

6. The optical sorting machine (1 ) according to one of claims 1†o 5, characterized in that the optical sorting machine (1 ) comprises an interface module

(1 .91 ) connecting the optical sorting machine (1 )†o the world-wide backbone network (interne†) (1.92), wherein data on the sorting modes uploaded†o a cloud-based database (1.921 ) by a plurality of sorting machines (1 ) is used†o assess the generated risk of contamination in the local area (1 .94) and/or is used for creating alerts†o other users in the geographical area (1.94).

7. The optical sorting machine (1 ) according to one of claims 1†o 6, characterized in that the firs† wavelength range (1.631 ) is between 450 - 495 nm in the blue light range and the second wavelength range (1 .632) is between 495 - 570 nm in the green light range.

8. The optical sorting machine (1 ) according to one of claims 1†o 7, characterized in that the one or more optical detectors (1 .7) a† leas† comprise cameras (1 .71 ) for splitting the received reflected or emitted light into two wavelength ranges using a prism with a dichroic filter element.

9. The optical sorting machine (1 ) according to one of claims 1†o 8, characterized in that the illumination means (1 .6) cause the objects (1.132) to fluoresce and comprise a† leas† one incident light emitter emitting one or more incident light beams, and further characterized in that the optical detectors (1 .7) are configured†o detect fluorescence emitted from the objects (1 .132).

10. The optical sorting machine (1 ) according to claim 9, characterized in that the incident light beam(s) comprise ultraviolet light (UV).

1 1. The optical sorting machine (1 ) according to one of claims 9 or 10, characterized in that the sorting machine (1 ) sensing the fluorescence emitted by the optical detectors (1 .7) comprises optical filters (1 .81 ) for isolating the incident light beam(s) and the emitted fluorescence.

12. The optical sorting machine (1 ) according to claim 1 1 , characterized in that the optical filters (1 .81 ) for isolating the incident light beam(s) and the emitted fluorescence comprise filter fluoromefers using filters†o isolate the incident light and fluorescence or specfrofluoromefers using diffraction grating monochromators†o isolate the incident light and fluorescence.

13. The optical sorting machine (1 ) according to one of claims 1 to 12, characterized in that the one or more optical defectors (1 .7) a† leas† comprise cameras (1 .71 ) .

14. The optical sorting machine (1 ) according to one of claims 1 to 13, characterized in that the sorting means (1 .5) comprise one or more ejectors (1 .5) ejecting undesired objects (1 .132) from the objects (1 .13) to be sorted, thereby providing the accept stream (1 .42) and the reject stream (1 .43).

15. The optical sorting machine (1 ) according to claim 14, characterized in that the one or more ejectors (1 .5) a† leas† comprise pneumatic ejectors (1 .51 ).

16. A method for an optical sorting machine (1 ) for sorting of granular objects (1 .13; P), which comprises passing the objects (1 .13; P) by a hopper (1 .1 ) and/or feeders (1 .1 1 ) and/or chutes (1 .12; 1 .121 , 1 .122.1 12x)†o a sorting zone (1 .4), the sorting zone (1 .4) comprising illumination means (1 .6; 3, 3'), one or more optical sensors (1 .7; 7) sensing reflected (1 .62) or emitted (1 .63) light (FE) from the objects (1 .13; P), and sorting means (1 .5) for separating desired objects (1 .131 )†o an accept stream (1 .42) and undesired objects (1 .132) to a reject stream (1 .43), characterized in that luminous intensities of the reflected (1 .62) and/or emitted light (1 .63) or a photometric equivalent thereof, sensed by the optical sensors (1 .7; 7) are measured by optical splitter means (1 .81 ), in at leas† a firs† and a second wavelength range (1 .631 , 1 .632. 1 .63x), in that trigger identifications (1 .221 ) of undesired objects (1 .132) are triggered by trigger means (1 .2) based on their luminous intensities measured in the at least first and the second wavelength range (1 .631 , 1 .632) of the emitted light (1 .63), by means of a plurality of defined sets of at least 2-dimensional intensity trigger vectors (1 -21 ), wherein each defined set of intensity trigger vectors (1 .21 ) defines an intensity color space (31.34) of trigger identifications (1 .22) of objects by luminous intensities in the at least first and second wavelength ranges (1 .631 /1 .632), wherein at least one of the intensity color spaces (32, 33) comprises trigger identifications (1 .22) of objects with a higher probability of correctly identifying undesired objects (1 .132) than at least one other of the intensity color spaces (31 ,34), in that a frequency or number of undesired objects (1 .132) is measured by a frequency counter (1 .31 ), which undesired objects (1 .132) are identified in the at least one of the intensity color spaces (32, 33) with the higher probability of triggering an undesired objects (1 .132), and an overall frequency or number of objects passing the sorting zone (1 .4) is measured by measuring means (1 .32), and in that the sorting means (1 .5) are operated and steered by a steering device (1 .9) comprising two switched modes switching between sorting and monitoring the operation of the sorting machine (1 ), wherein if a predefined threshold value of the frequency or number of undesired objects (1 .132) identified in the at least one of the intensity color spaces (32, 33) with the higher probability as a portion of the measured overall frequency or number of objects passing the sorting zone (1 .4) is exceeded, then the sorting means (1 .5) are switched on to sorting mode (1 .931 ), while otherwise the sorting means (1 .5) are switched off to only monitoring mode (1 .932) of the optical sorting machine (1 ).