Automatic pin bone cutting for cod, saithe, haddock and salmon
Summary of the project's final reporting
Improving bone detection on soft bones
Valka has obtained very good results in detecting pin bones and belly flap bones in white fish. Belly flap bones can have very different shape from the pin bones and need special handling to identify in the analyses software. Neck bones can show up in fillets due to filleting defects that occur in the heading or filleting machines and have been missed by the pre-trimming operators. These bones are now accurately detected in the x-ray machine and either the entire fillet is rejected or only the first loin piece.
Bones in salmon are however more difficult to detect. In x-ray images the clarity of the pin bones vary significantly. The end of the bones are difficult to detect and future research will be guided towards making a focused search at the end of detected bones such that they will not be detected shorter than they are. The images get clearer when the x-ray machine is operated at lower speed and it might be necessary to run a salmon cutting machine at a lower speed than the normal operating speed of 460 mm/s.
Obtaining the 3D position of bones
NTNU has now completed implementing an algorithm that can accurately obtain the 3D position of pins in test pieces that have been constructed. Valka has now ported this code into the Valka code base and have obtained excellent results on the test pieces that were constructed.
There are many special considerations that need to be evaluated when 3D position of bones from x-ray images of fish fillets is to be reconstructed. Valka is still working on this development and estimates that several months of work is remaining before an accurate 3D reconstruction of bones will be up and running.
Once the 3D position has been obtained it is of utmost importance to be able to cut out the bones at an angle with maximum accuracy. The initial tests with Valka’s new robot shows that it is by far superior to the earlier design and it looks as it will meet the speed and accuracy requirements needed.
Testing of new methods with fish from various parts of Norway
During the last year the machine has been tried in tests and under commercial circumstances on various quality’s and age of raw material both in Norway and in Iceland. Fish from the Barent sea has been processed in significant quantity in the Samherji´s UA plant and the fish processed in Gryllefjord has been a mix of seasonal coastal fish from Troms and Finnmark. Additionally, trawl fish from deeper waters around Norway has been processed in Gryllefjord.
The main conclusion is that the age of the raw material, time of the year, rigor state or fishing ground has no direct effect on the cutting and bone detection. However, since the general quality of the raw material is affected by all these factors, they have an indirect effect on the performance of the cutting machine. With increased age and poor handling, the fillet loses some of its elasticity and becomes softer, when this occurs the bones get loser in the flesh. This can result in a few issues, the first is that some bones can move out of their ‘normal’ position during handling and production. That means that after filleting, skinning and pre-trimming some bones in the fillet are no longer laying parallel to the moving direction of the belt but can be in any position including perpendicular to the moving direction of the belt. Bones in that position are harder to detect, especially when running at the normal operating speed of 460 mm/s. Another issue is that the pin bone yield has a tendency to be worse when the fish is soft, this happens when some of the bones ‘float’ out of position making the surface area of the bone region bigger. The third main issue is that gaping, a flaw often correlated with old and soft material has negative effect on weight estimation and the cutting accuracy of the portions.
Definition of a complete processing line
A complete processing line has been working in the Samherji´s ÚA plant for close to a year. The line is working well and the cutting machine gives excellent flexibility and is able to make all the cuts that have been requested with very good accuracy. There are never the less several issues have caused some challenges. One is that it is difficult to ensure that the portions are always fully separated. This is especially the case on softer fish and if ligaments or membranes are not removed in the pre-trimming. This is not an issue when portions are manually graded after the cutting machine but is really critical when the portions are graded automatically. A good combination of cutting speed, orifice size and pressure has been identified that gives very good results for most sizes of fish. Some final adjustment is still needed on the biggest fillets where ligament and membranes have more effect.
In the UA plant Valka delivered for the first time an x-ray machine behind the cutting machine such that pieces that have any remaining bones can be rejected and the bones are then manually removed from those portions.
The portion grader is working well when all the portions have been properly separated in the cutter and the grader can accurately grade portions down to 60–70 mm in length and up to 3 parallel portions on the belt from the same fillet. The key benefit of the Valka grader is that it can maintain the alignment and orientation of the portions and they can therefore be fed automatically to IQF freezers, vacuum packing machines or the Valka Aligner grader.
There are two main issues that Valka sees for improvement. One is to have a computer monitor such that the person taking out the extra bones can see on an image where the bones are. Without the image it can be difficult to find small bones and they might thus be missed. The second issue is that it would be beneficial to be able to ensure that there are no extra bones in the fillets when they are fed into the cutting machine. That is other bones than the belly bone and the pin bones. To solve this Valka plans to have an x-ray machine as well as quality analyzer unit in front of the trimming line and mount as well computer monitors on the trimming line.
When designing a complete processing line it is important that the line can handle the capacity needed. For bigger processing plants a dual lane cutter is a very good solution as it saves significant processing space and makes the flow in the plant more efficient.
Formation and elimination of white material on salmon fillets
Appearance of white material is a challenge for water-jet cut salmon fillets as the sensory quality is compromised. However, with certain combinations of operational parameters, like the cutting orifices, pressure, cutting speed and raw material treatment, it is possible to counteract this effect. Valka has made good progress in a solution for automatic cutting of pre-rigor salmon fillets, and water-jet cutting of salmon fillets can be a promising future technology for the salmon processing industry. Nofima as well as personnel from Bremnes Seashore worked with Valka on this part of the project.
Valka has in this project made very significant progress in developing a new processing line that can increase the competitiveness of fillet production in highly developed countries like Norway. With the Valka Cutter it is possible to make any type of cuts—including what some in the industry call ‘crazy Chinese cuts’ due to the complexity—and those include cutting out the belly flap bone and splitting loins or other portions lengthwise. This is needed to be able to be able to meet the strict requirements needed for the retail market which gives fish processors great opportunity to produce added value products in their fillet production. Valkas hope is that this project will assist in building up a strong fillet production industry in Norway in the near future.
Valka started in 2009 to work on automatic cutting of pin bones from fish fillets. Valka engaged in a development project with HB Grandi, Samherji and Ný-fiskur. The aim in the first project was set on vertical cut based on 2D analyses. The project was supported by the AVS research fund. An x-ray guided cutting machine for small redfish based on technology from Valka was delivered to HB Grandi in August 2012. That machine has now been up and running for about one and a half year operating at the accuracy and throughput promised. What primary remains in that project is to remove the fillet section which contains the bones automatically but that is being done manually in the current line.
On the basis of these promising results it was decided to take the project to the next level such that the machine could cut bones from cod, saithe, haddock and salmon fillets of any size. Early in the year 2013 it was decided to make a test on the machine in HB Grandi in Reykjavík in cutting cod fillets with vertical cut. The results gave better results than expected. It turned out that the yield loss when the pin bones are cut out of cod fillets of various sizes ranges from 5–7 per cent which is quite compatible to the yield loss obtained with manual cutting. HB Grandi decided then to go ahead and purchase a machine for cutting cod fillets and that machine was installed in August 2013. One similar machine was in the beginning of 2014 sold to Gryllefjord Seafood AS in Norway.
At this point Valka has developed a prototype of the system where the pin bones are removed by using 2D shadow picture. The next step is to extend the technique for cod, haddock, pollock and other fish with sloping pin bones. This can be done by generating 3D picture, instead of 2D shadow picture, for detection of the pin bones. The 3D information of the position of the pin bones will be used to cut out the pin bones with maximum utilization of the fish fillet. This can only be achieved using cutting pattern that can follow the slope of the pin bones. The machines already sold can be upgraded to perform angular cut and thus reduce the yield loss.
To be able to cut pin bones automatically from cod, saithe, haddock and salmon with better accuracy than can be done manually today.
With this system it is possible to increase the fillet utilization and gain more accuracy in pin bones removal than is currently known in the industry. The cutting robot will also be used to cut the fillet into desirable pieces, with maximum utilization in mind. It is estimated that utilization of the fish fillets can be increased by 1.5–2.5 per cent for cod and 1–2 per cent for pollock and haddock.
For cutting large fillets with acceptable accuracy it is necessary to measure the 3D position of the bones and use angular cut.
An automatic pin bone cutting machine will revolutionize how fillet production is organized. Defining a new processing concept around this new technology is therefore of great importance as well.
Today the most common cutting pattern focuses on maximizing the size of the loins and such cuts cannot be made with traditional portioning machines found in the fishing industry. An important aspect is as well more efficient portioning.
The project is split into six parts:
1. Improving the bone detection on soft bones and partly soft bones as are commonly found in cod and salmon.
2. Utilizing stereo vision techniques to obtain the 3D position of bones with maximum accuracy and defining optimum cutting paths around the bones based on capability of a cutting robot with multiple degrees of freedom.
3. Testing of the methods above with fish from various part of Norway in various conditions – most importantly both pre-rigor and post-rigor.
4. Definition of a complete processing line built on this new technology which will make Norwegian fillet production more competitive in the world market. The economical aspect of the new processing technology will be evaluated as well.
5. Identify causes for formation of white material when salmon fillets are cut with waterjet cutting technology and test different alternatives to eliminate or reduce this formation.
6. Automation of sorting of portions after cutting.
Valka is the project owner and is leading the project. Valka will hold the primary responsibility for all development work in the project. As well Valka will supervise the work done for individual tasks in co-operation with other project members. Several people from Valka will have input into various tasks.
The results will be described in a project report and by presentation on a workshop arranged by FHF.