Smart farming | Spectral Engines

SmartFarm - concepts and collaboration for better future farms

As smart farming and digitalization are making their way to livestock farms, spectral sensors are becoming an important factor in increasing the productivity of farms and the welfare of animals.

In late 2017 Spectral Engines partnered with Technical Research Centre of Finland (VTT) and Natural Resources Institute Finland (LUKE), two of Finland’s leading institutions in the area of bio economy digitalization for a project called SmartFarm (Smart Sensors and concepts for farm digitalization). The project aims to speed up new ways of working and the wellbeing of animals at dairy farms based on digitalization and developing next-generation sensor technology. Other companies collaborating in the project are Faba, M-Tech Solutions Oy, GrainSense Oy and Pehutec. Two scientists of the project, Dr. Matti Pastell1 and Dr. Ben Aernouts2, open up the background of the project and its main goals and benefits for the agriculture sector in the future.


The two scientists met last year at the EAAP conference in Tallinn. Their strong overlap in interests and research objectives got them discussing about future collaborations, and from that moment on, they started preparing Dr. Aernouts’ research visit at Luke Helsinki for the first half of 2018.

As the leading innovator of sensor technology, Spectral Engines was an easy choice to partner up with in the project as well.


“The project and collaboration with Spectral Engines was initiated by VTT. They have a long history in developing optical sensors and had the idea of using them for milk and silage composition, and after I was contacted by Dr. Mikko Utriainen, I was immediately interested in the idea and opportunities that new sensors bring to agriculture. After that we developed the project proposal together”, Dr. Pastell recalls the initial stages of the project.


Science for happy cows

As an offspring of a medium-sized dairy farm in the north of Belgium, Dr. Aernouts realized from a young age that efficient milk production can only be achieved with healthy and happy cows.

“Animals need to be carefully monitored to get insight in their health status. Because the number of farmers has clearly decreased over the last decades and the demand for milk products has slightly increased, a single dairy farmer is managing much more cows today compared to a few years ago. Luckily, partial or full automation of the feeding, milking and cleaning processes have helped them to manage their workloads. Nevertheless, the monitoring of cow health and behavior still requires a lot of time. Therefore, additional help from accurate, reliable and robust sensing technology can significantly support the farmers in optimizing herd management.”


According to Dr. Aernouts, milk is a good reference point for the overall wellbeing of the animals.

“Milk production involves a very intensive interaction with blood circulation and thus can be used as a mirror for the animals’ metabolism. By regularly analyzing the amount and quality of the produced milk, the cows’ health can be monitored without affecting their welfare. As such, early detection of a health issue, followed by early treatment, can reduce the severity, milk loss, medication costs and the risk of early culling and secondary diseases. Moreover, it can limit the spread of infectious pathogens in the herd and reduce food waste through the separation of spoiled milk. Individual monitoring of metabolic and health status of the cows can aid to improve profitability and ultimately lead to more sustainable dairy production.”


Dr. Pastell is on the same track and emphasizes that developing accurate methods for real time measurement of animal feeds and manure are important in order to improve resource efficiency.

“Fast on-farm options for measuring milk and silage composition enable more accurate feeding of cows which leads to improved productivity. Daily milk composition data can also potentially be used to monitor the cow’s energy balance and ensuring better health. Measuring the nitrogen and phosphorous content from liquid manure during spreading would enable using the nutrients more efficiently leading also to a lesser environmental impact.”



Towards the next step

At this stage of the project, VTT is preparing a sensor prototype for on-farm milk quality analyses. Since Dr. Aernouts has already gained a lot of expertise in autonomous on-farm measurements of milk quality with NIR, he has helped VTT with the design of their sensor. During the next stages of the project Dr. Aernouts will be helping with the implementation processes.


“In a completely autonomous system, the sensor needs to communicate with the milking equipment. For example, information on the progress of the milking or rinsing process is essential to initialize the intake of the milk sample or cleaning solution by the sensor. On the other hand, information on the measured milk quality can be sent back to the management software – which is typically linked with the milking equipment – to couple it with the respective cow and milking identifiers. Finally, the milk quality data needs to be translated into information on cow health and performance relevant for the cow and farm management. To this end, we believe that each individual cow needs to be monitored in relation to her own reference, in terms of history, and to the herd she is part of. One of the goals of my research stay at Luke is to collaborate with Dr. Pastell on the development of tools for time-series monitoring of animal health and behavior.”


Correct sensor data leads to correct interpretation

Both scientists have noticed in their work that sensor technology typically produces a lot of data of which the majority is irrelevant for the farmer. In an ideal situation, the farmer has continuous access to an up-to-date, clear and accurate overview of the actual status of his animals without spending time on interpreting the sensor data. This monitoring system should give a warning when a change in an individual animal, a group or an entire herd is noticed. This requires correct interpretation of the collected data and the extraction of reliable information using advanced data-processing techniques. One of SmartFarm’s main aims is to demonstrate the value of combining data from multiple sources to a unified analysis to make the most out of digitalization in farm management.


“For this reason, our research also focuses on the development, validation and implementation of intelligent data-algorithms which provide explicit information on the cow’s health, welfare, efficiency and environmental impact through interpreting and combining different sources of sensor data and relying on physiology-based knowledge. Today, a lot of technology is already used in agriculture and farmers typically believe that the sensor data is correct and that there is a direct link with the process of their interest. However, this is often not true: the obtained signals are influenced by different sources of noise and disturbance, which can be related to imperfections of the technology and the complexity of the monitored biological processes. I am convinced that solving these issues will result in major progresses for precision agriculture. The quality of the measured data will improve due to better technology in combination with continuous validation and recalibration.”


Facing the challenges with innovative sensor technology

Dr. Aernouts believes that fundamental research together with a combination of different sources of data will result in better understanding of the obtained measurements. According to him, one of the major challenges for on-site, on-line and in-line monitoring of product quality is typically that the size of the technology is related to its performance. Spectral Engines’ sensors seem to be an excellent way to tackle these challenges. Dr. Aernouts was already acquainted with Spectral Engines products before joining the SmartFarm project, as the university of Ku Leuven has two Spectral Engines’ modules, each for a different wavelength range, available in their laboratory.


“For easy implementation and reduction of costs, miniaturization of the sensors is desirable. However, this usually results in a reduced accuracy of the obtained measurements. I am convinced that the innovative technology used in the Spectral Engines sensors combine the potential for both a small design and high performance. I am looking forward to figure this out throughout the project.”

Dr. Pastell agrees on the performance of Spectral Engines’ sensors and is looking forward to the rest of the project and on-farm pilots.


“The project has progressed in schedule and we have done pilot measurements and experimental set-ups for the new Spectral Engines modules with both silage and milk. The results of measuring the composition of both materials are promising.  Spectral Engines sensors are well suited for in-line milk measurements and measuring silage dry matter content. The small size of the sensors together with high sensitivity makes them a good fit for practical on farm use. Low cost is also an important factor in agricultural products. The data from the sensors will be integrated with other farm data via a cloud solution in the farm pilots and visualized to the user via a single user interface. We will continue to develop calibration models with more calibration samples and will have an on-farm pilot running next year”, Dr. Pastell concludes.


If you’d like to know more about our solutions for smart farming, don’t hesitate to contact our sales and business development.


1)Dr. Matti Pastell is the Associate Professor and Research Manager at Luke and the leader of Luke’s part of the project, and his main task is to develop the data model for combining data from different sources and develop a data-based farm management pilot in the project. He has worked in the area of Precision Livestock Farming for more than 10 years, and his main area of research has been in automatic health and welfare monitoring of livestock. Currently Dr. Pastell is working on developing methods for measuring cattle behavior using indoor positioning systems together with accelerometers and improved models for detecting abnormal animal behavior.

2)Dr. Ben Aernouts has a PhD degree in Bioscience Engineering, and during his PhD research he studied the optical behavior of food. Later he has used the insights obtained in his former research to improve the design of optical sensors for food quality monitoring. The focus of this work was on the development and validation of on-farm and in-line sensors to monitor milk quality and cow health. Since October 2016, Dr. Aernouts has been the assistant professor of Management in Livestock Production at the KU Leuven University in Belgium. Besides his teaching activities, he is building a research group working on Livestock Technology with a focus on the development, implementation and validation of innovative sensor technology and data-processing algorithms to support animal management in livestock production.