Optical sorting: a brighter, better, and clearer view
By Dr. Heike Knörzer, Petkus, Germany
The future of seed and grain sorting is a bright light. Is it? Indeed, advanced optical sorting systems with divergent optics and illumination techniques are a booming trend. The range of optical sorting systems is increasing and more and more vendors are on the market. Sometimes, general promises such as steady sorting accuracy levels of >99 percent are proclaimed. Theoretically it is possible. Often it is also practically possible. But as nature is highly variable and never constant, accuracy levels for all lots or sorting tasks of 99 percent are not realistic. But the fact is, that optical sorting can rescue production and prevent tremendous revenue loss.
In certified oat seed production in Europe, there is a zero tolerance for wild oat. A Spanish oat seed producer faced the issue of having 5714 kernels of wild oat in one tonne of mechanically cleaned oat seed (0.02 percent). The official certificate would have been labelled with "not certified", as his own 5.5 kg sample analysis with 31 wild oat kernels indicated.
Ten hectares of seed production most probably lost and to be sold as commodity. With an average oat yield of two t/ha and a commodity price on the CBoT stock exchange with €250/t, the difference in revenue between commodity oat and seed oat would have been €5,600; excluding the higher price of the field production. With a yield of 5 t/ha, the difference would be €14,000.
The sample was sent to the Roeber Institut GmbH, a member of the Petkus Group, in Wutha-Farnroda, Thuringia, Germany, where it was analysed before and after optical sorting with the Petkus OS 901 colour sorter with one re-sort cycle. Wild oat could be completely separated by the OS 901. The "accept" fraction showed a purity level of > 99 percent. There were no kernels of wild oat left in the sample after sorting.
According to the Department of Primary Industries and Regional Development"s Agriculture and Food Division (DPIRD) in Western Australia, the annual costs to the industry of wild oats was estimated to be USD 80 million, and wild oats having the potential to decrease yields by up to 80 percent.
Sorting results above the standard
Optical sorters are doing a great job where mechanical cleaning reaches its limits or a specific or high quality is needed. In addition, weather conditions or cultivation flaws can cause field production failure which can only be offset in some way by optical sorters. Roeber can give vast examples for those rescue operations:
- Deoxynivalenol (DON) level reduction due to fusarium infestations to below threshold limits of 200 µg/kg for baby food purpose;
- Separating paddy rice kernels with a red coloured bran from white rice (reduction from 13% to 0.11% contamination);
- Sorting a rye lot with 15 percent contamination of triticale to final 0.25 percent contamination;
- Separating durum from soft wheat for seed or commodity pasta production;
- Removing off-types of sweet corn originating from commercial sweet corn volunteers with a final purity level of target sweet corn grains of > 99 percent;
- Counting and analysing statistically the nature and amounts of defects and impurities in Arabica coffee beans such as sticks, cherries, black seeds, pergamino, fermented seeds etc.
The list for optical sorting applications is long. The machines became an integral part of the seed and grain processing industry. Optical sorters are often considered as easy to operate machines which can do all sorting jobs easily: colour, shape, seed size and ingredient detection and separation. Nevertheless, it all depends on the composition and the material to be sorted.
Sorting results can differ substantially in purity levels between crops, but also between varieties. "You have to be aware that results can differ even within crop lots or for all variants," says Dr Kai-Uwe Vieth, Sales and Development Manager at Roeber Institut GmbH, knowing about the influence of input contamination level, required ratio of good material in the reject and differing characteristic traits of varieties.
"Living organisms are too variable per se." In addition, high sorting quality or purity is at the expense of good seed loss in the reject fraction. The last one percent of purity costs "accept" kernels.
The development goes on
"Obviously, the sorting results of some optical sorters on the market are above the standard compared to common cleaning technologies", says Vieth. "But that doesn"t implicate that the development is at its end." That is the reason why in the Petkus Group, a dedicated team is continuously evaluating optical components to scrutinise their own technologies, whether they are the benchmark or how they can be further improved. "We are in permanent exchange and partnership with scientific institutes in order to future-proof our optical sorter portfolio", says Vieth.
The market is constantly screened for advanced components, and cameras are tested in the Roeber optical laboratory in Karlsruhe, Germany, that is not only associated but also locally in close touch with nearby research and development partners. The Karlsruhe office successfully applied for the "Karlsruher Business Innovation Program." Within this "Innovation Program," the office for economic development of the city of Karlsruhe cooperates with the European Organization for Sustainable Development GmbH (EOSD).
High-resolution cameras and long-life LEDs alone are no assurance for good sorting performance. The synchronisation of the optical system, the hardware-software package as well as the design of the optical boxes and the ejection play an essential role when it comes to clear recognition of defects. On the hardware side, it means that the sensitivity of the cameras and the power of the illumination have to be sufficient.
Regarding software, the deep or machine learning offers new possibilities, and there is a steady development noticeable. Big data analysis is especially for optical sorters a great chance for improved software programs. Automatic teach-in programs help operators to adjust the machine more easily to different sorting tasks. The easier it looks for the operator, the more complicated it is in the background and the more knowledge is needed with regard to the evaluation and validation of algorithms.
"When I compare our first TeachNSort software package with the actual TeachNSort Smart+, there was a lot of work done and improvements for the benefit of our clients and the sorting accuracy", says Vieth.
Camera resolution is widely discussed and is considered to give information about sorting accuracy. "That is true to a certain degree", says Vieth. "But without appropriate object splitting algorithms, a single seed analysis during high capacity operation is weak, just to give an example." Knowing about the details and the pitfalls of optical sorting systems, Vieth is sure about the Petkus/Roeber optical sorter series "OS", that the package is the most sophisticated and complete.
Optical sorting systems
Another key component is the illumination. The wavelength, the spectra, the power of the light source and from where the light is sent make the difference. This range of illumination characteristics offers a more precise selection of illumination system for individual sorting tasks.
For example, often it is impossible to separate sunflower from sclerotia by using standard RGB systems. Therefore, a NIR system is recommended. To separate red rice from white rice in paddy production, where the red coloured bran is beneath the rice husk and can only be seen after husking, a powerful translucence system is needed. A combination of RGB and shape could be successfully applied where two different soybean varieties with different hilum colour and seed dimensions have to be separated.
Optical sorting can be even more precise than the human eye when defect type and illumination systems fit to each other. Absorbed, reflected and/or transmitted light can provide the information that is needed for defect detection. Light source, optic design and machine construction decide which illumination can be used in respective optical sorters.
When results with shape and wavelength in the visible range of 400 – 700 nm (RGB) become limited, then the OS 901i with NIR/SWIR camera technology can be a solution that has a 950 – 1,700 nm wavelength bandwidth. The state of the art RGB and NIR technologies can only detect defects on the seed surface. But when defects in the inside of the kernel have to be detected and the kernel has some translucence properties, than the Petkus/Roeber OS 901 t comes into play. The strong, homogeneous and high quality LEDs inside the OS 901 t as well as the special light configuration allows for a look through some grain species. Only the Petkus/Roeber OS 901 can offer variable luminance for variable defects, combined in one platform.
OS f-class as an example for new optical sorters
A new type of optical sorter is launched by Petkus/Roeber in 2018. The OS f-class is the first in optical sorting for small seeds such as vegetables, flowers, grass and spices. It allows cleaning of small seeds to the highest quality level where nowadays-standard colour sorters fail due to unprecise ejection of standard air nozzles, low resolution or insufficient particle orientation. Excellent sorting results are even possible for small seeds like Origanum vulgare (TKW ~ 0.1 – 0.13 g).
Intensive testing with various species was done in the Roeber optical test centre with impressively high results. Inter alia, a client conducted comparative tests with the OS f-class and the client"s existing machine. The results were analysed by the client himself. Tested species had a diameter of around 0.5 mm. The proportion of defects in the accept fraction was reduced by a factor of two by simultaneously doubling the throughput related to sorting area. Compared to what is possible with existing machines in that area of application "the precision is very exciting", concluded the trial team around Vieth.
In the OS f-class, seeds are transported from the feeder along a stainless steel chute down to a specially designed profile belt system to be scanned at the end and ejected nearly immediately behind the inspection line. The belt aligns the seeds to the ejector nozzles to ensure precise positioning. Custom designed by Roeber, the OS f-class has got a new ejector-nozzle technology with an incredibly narrow pitch. "You could put the nozzle block in your pocket", says Vieth.
Because of the special feeding system, the seeds can be round and tumbling or leafy or arbitrarily shaped. In addition, the density of the product can be high or very low. This makes the sorting system applicable to an enormous range of products and defects.
A high-resolution full colour camera system based on the latest high-tech components detects tiny differences to ensure the purest seed lots. The product is not only scanned by a high-resolution camera but also embedded into an innovative light and optics combination. The high contrast allows the sorter to distinguish tiny differences.
Details of the small seeds are easy to see and can be easily evaluated on the large 19" touchscreen. Over internet support, and by using built-in LAN, the sorter has the capability of true remote control due to real image analysis.
"An important point is that Petkus can give effective remote support and help with product set-up because the company can see the real pictures of the client"s product on the screen. Sorters that don"t have the real imaging have limited possibilities for remote support", says Vieth.
The complete TeachNSort software can also be installed on customer"s office PC. It has the advantage of just taking the images of good seeds and if necessary also bad objects at the OS f-class or OS 901. The data are easily exported to a USB stick and imported at the office PC. Then the user can perform the teach-in while enjoying a cup of coffee and while the OS f-class or OS 901 is working in production. This makes the whole teach-in process more comfortable and more efficient.