Robot vision applications – viewpoint from automation in the fish sector

Citation

Ove Østvik, S. (2011), "Robot vision applications – viewpoint from automation in the fish sector", Industrial Robot, Vol. 38 No. 4. https://doi.org/10.1108/ir.2011.04938daa.001

Robot vision applications – viewpoint from automation in the fish sector

Article Type: Viewpoint From: Industrial Robot: An International Journal, Volume 38, Issue 4

Automation in the fish-processing industry has a good track record in mechanized automation using machines with some degree of flexibility to adapt to the fish as the processing object. The traditional Baader fish-filleting machines, introduced in the 1950s, are probably the best examples of sophisticated machines that have had enormous significance for fish-processing industry in the last century – substantially reducing the need for manual labor. Further advances in fish-processing automation include the introduction in the early 1990s of the first 3D measurement-based portion cutting machines. Some years later, in the 2000s, the first vision-based fillet-trimming machines were launched. The trend in the fish-processing industry is that it wants to automate regular processes as much as possible.

Despite the need and desire for automation of processes and operations in the fish-processing industry, achieving high degree automation has turned out to be quite challenging – for several reasons. In fish (and meat) processing, the objects to be handled and processed will normally have substantial variations in shape, color, texture and other external and internal properties. This presents a challenge for any attempt to automate a process in this industry. Furthermore, the market trend is to focus on safe food, predictable product appearance and product diversification. Thus, it is challenging to automate specific processes, due to the variability in the product, in one side, and degree of accuracy, flexibility and speed required by the current market.

Perhaps, due to the challenges of automating even specific single processes that are currently carried out by manual labor, the automation of processes is seen as a way of replacing specific manual operations. A traditional approach to development of automation technology is to focus on single specialist machines, combining such machines into processing line and reducing the number of processing stations requiring manual labor. This approach to automation is not necessarily the future for automation in the fish-processing industry. The possibilities offered by internal and external 3D vision and detection beyond human senses, implies that technology can be used to create new processes and products – processes and products not previously imagined within the framework of involving manual labor. Furthermore, using more flexible robot technology, the processing system can be more integrated and holistic, with flexibility to add multiple processing capabilities to a single holistic system.

Everything is possible – a cliché, but it might be a good lead, and soon a reality, regarding automation of complex operations in the fish-processing industry – even using currently available technology hardware and software as building blocks. The greatest challenge is not necessarily on developing the building blocks, but more on system understanding and system integration seen in perspective. The technology development process towards future automation needs to include in-depth knowledge of the target industry and its needs. This involves understanding the entire process from bulk raw material to finished products, while also understanding the market dynamics and requirements. The automation systems of the future should be developed based on this holistic view of the available technologies, raw materials, end products and their markets. The development process for such systems involves selection of technology elements for machine vision and sensors, and technological tools for robotics, end effectors and mechanical solutions. Additionally, a systematic product development process, using flexible hardware and software platforms, requires a significant amount of creativity during software programming and hardware integration. An effective project arena for such a development process should include the dedicated technology supplier(s), sufficient R&D resources and technology users (e.g. the fish-processing companies).

An example of a recent automation solution is the automated slaughter line for salmon presented in current journal issue. For this application, a 3D laser triangulation system is used as vision input. Although 3D laser triangulation is a well-established technique, the challenges of non-singulated, overlapping and twitching fish made this a non-trivial application to solve. This is apparently basic development but the experience is that development of such applications involves a certain degree of creativity and discovery of specific novel solutions and approaches. The development of this slaughtering line is ongoing, and will result in a system that combines the slaughtering operations with more functionality, such as quality-based sorting of the fish.

An example of a successful automation demonstration by SINTEF involved automatic and adaptive trimming of salmon fillets, using accurate and completely object-adaptive 3D cutting with a multi-tool oscillating knife. In the SINTEF Sealab laboratories, we demonstrated a rapid development process with a multi-disciplinary team. Within a period of only a few months, we had set-up a 3D machine vision system, developed a multi-tool end effector, and developed a vacuum-suction pallet and transport system for the fillets. This demonstration served to illustrate how even a very challenging task can be solved in a short amount of time using a multi-disciplinary team with a holistic view of the problem to be solved. The purpose was not to develop a commercial solution, but to demonstrate a feasible development approach for such a complex problem. Around the same time, and later, two technology companies have developed and launched vision-based fillet-trimming machines.

Figure 1 Example perspective over future automated fish-processing line based on: detection by multifunctional machine vision, standard flexible robots with special adapted tools, logistic system to combine all operation to holistic system and repeated detection for automated feedback which enables self-learning abilities of the system

From the above two examples, and others, we have learnt several key points and relevant trends for the future of automation in the fish sector:

• •

  Multi-sensor systems will merge vision-based sensors with other types of sensors, enabling complete multi-modal sensing of the product.

• •

  There will be a trend towards holistic processing systems, with a transition from the specialized single-operation machines of today to multifunctional, reconfigurable and adaptive processing systems.

• •

  The setup and configuration of future fish-processing automation solutions will be simplified, due to an increase in the level of self-learning and built-in intelligence – enabling the reconfiguration to, e.g. a new species or new processing task to be done on the factory floor, instead of through long research and development projects.

We believe that most fish processing can be almost completely automated. Our experience so far, as demonstrated in commercial solutions and technology demonstrators, has upheld this belief. Rapid and continuous growth in the industrially available technology elements, and flexible software development platforms, enables a reduction in the amount of time from conceptualized idea to an implemented solution. In addition to this increase in the amount of easy-to-build-with hardware and software building blocks, we see that the tasks in the fish sector (and food sector, in general) are challenging enough to be on or beyond the frontier of what is possible using only existing building blocks. Owing to the nature of automation challenges in the fish sector, the development of novel methods, techniques and approaches in this sector can benefit and further advance the field of industrial robotics, in general. Thus, the future of advanced robot-based industrial automation is a bright one, not only in the fish sector (Figure 1).

The author

Stein Ove ØstvikResearch Manager, SINTEF Fisheries and Aquaculture, Trondheim, Norway

Further reading: www.sintef.no/Fiskeri-og-havbruk-AS/ProsessTeknologi/