2 replies to this topic

[deepti]

Hi, We are using optical sorters in our factory to inspect Salad Leaves. The problem is we are getting constant complaints of FOBs in the finished product. The factory is separated physically between Low Risk, High Care and Final Pack. There are no chances of cross contamination from low risk to High care products. The foreign objects such as stones, feather, insects and plastic etc. which would have originated from the farm in most cases are found in finished product. What are your views on this situation? Can someone please let me know what kind of test pieces are you using for testing the sensitivity during production and where to find them? Thanks Deepti

[Steve_T]

Hi Deepti

 

Based on past experience with optical sorters, their use is limited to sorting by colour differentiation.  So for example, the removal of dark coloured bruises, stems or stalks from a lighter coloured product (for example dark blemishes from diced carrots, apricots etc).  Conversely, the reverse is also true, removing light coloured blemishes from a predominantly darker coloured product (eg the white flecks of damaged almonds).

 

These days, many high-end optical sorters also include an x-ray function, to enable the sorting / removal of product based on density differences.  This heightens their effectiveness of foreign object removal in case the optical cameras are unable to differentiate by colour alone.

 

Optical sorters are usually calibrated by the actual foreign objects you are trying to remove.  In other words, the smallest object you would expect to find is run across the machine, and the cameras adjusted to the desired resolution relative to the speed of the product travelling across the "void".

 

However, as good as these machines are, it becomes much more difficult for the machine to operate as you might expect while the colour of foreign objects match more closely to the product you are trying to run.  For example, you may even find the lights that illuminate the product as it passes the cameras need to be changed to a different wavelength in order for the cameras to best detect the particular foreign objects you are interested in.

 

And being salad leaves, what is to stop the foreign objects being hidden from view of the camera by the leafy nature of the product?

 

I would be concentrating on trying to find ways to improve the washing cycle at the front-end of your process.  In other words, remove the foreign objects to begin with, and allow the optical sorter to operate as a quality inspection point.

 

Good Luck!

[jordanjgunn]

As was mentioned, optical sorting based on color differentiation alone has very real limitations with respect to overall efficiency and accuracy.  In the several years since this post was made optical sorters have been moving to upgrade their hardware configurations to include sensors within different ranges, most notably in the NIR range of ~900-1700nm.  Simultaneously, machine learning software companies such as ImagoAI have been developing algorithms to increase the accuracy and add additional parameters (e.g. moisture, mycotoxins, TSS, dry matter, internal defects, titrable acidity, etc.) to optical sorting systems.  This allows for a much more robust analysis and the successful separation of accepts and rejects, especially in instances where you're unable to differentiate because of similar colors.