PLF case results

Monitoring area preference of fattening pigs - September 2019

An algorithm based on conventional image processing techniques was developed to locate and distinguish standing and lying pigs on depth videos recorded by D415 cameras. The algorithm was evaluated using videos taken in a pig barn under field conditions, and preliminary results showed that classification accuracy of 98% could be achieved. The algorithm output allowed investigating the relationship between the area preference of pigs and potential influencing factors such as ambient temperature.

iSense at EAAP 2019 - August 2019

ISense attended 70th Annual Meeting of the European Federation of Animal Science (EAAP 2019). At the congress we introduced the camera system for monitoring area preference of fattening pigs. Some  field test results and practical perspectives were presented and discussed.

Promising performance of the ammonia monitoring system - July 2019

The developed gas measurement system was tested over a prolonged period of time for monitoring ammonia concentration in a pig barn under practical conditions. The system was capable of performing continuous measurement in real-time with sub-ppm accuracy. Please read the published paper for more details.

Further improvements have been made to enable multichannel sampling, scheduled tube flushing and zero-calibration. These additional functionalities are currently tested under field conditions

Field test - 3D camera in a fattening pig barn - October 2018

Multiple Intel® RealSense™ D415 cameras have been installed in a fattening pig barn for field testing. The cameras are mounted 3m above the floor facing straight down, and each camera monitors two 2m×2m pens of loose-housed fattening pigs. An industrial PC is used to pre-process and store the video recordings. The cameras have been operating for one month and the field test will continue indefinitely to investigate the durability of the set-up. The collected recordings thus far show that visually the location and the standing/lying position of the pigs are relatively easy to identify on the false coloured depth images. In particular, depth provides a more distinguishable indication of lying behaviour than colour images when pigs lie prone. As the IR-based depth measurement by D415 is independent on ambient lighting, it is also more suited for 24-hour continuous monitoring than RGB cameras. The further step will focus on automating the identification of standing/lying behaviour and the area of preference.

Colour (top) and depth (bottom) recordings taken three weeks after installation. The magnitudes of different depth are visualized with false colour.

Monitoring pig behaviour with 3D camera – May 2018

Intel® has recently released a new series of low-cost 3D cameras, i.e. RealSense™ D415 and D435. With state-of-the-art technologies the new models have improved depth sensing accuracy and are capable of adapting to a wider range of lighting conditions both indoors and outdoors. The new product can be an interesting candidate for vision based tracking and behaviour monitoring. ISense will soon start exploring the possibility of using the RealSense™ D415 camera for automated behaviour monitoring in fattening pigs.

ISense in COST Action CA16106 – April 2018

The cost of measurement is one of the major bottlenecks in the research of NH3 emission in livestock farming. In a COST Action training school, validation results of the NH3 measuring system developed within the iSense project was presented. Researchers from other institutes showed great interests in the work of iSense, and were considering using the same analyser for their measurements.

(for more information about the COST Action, see

Development of a portable gas measurement system – update March 2018

A cost-effective portable measurement system for gaseous NH3 and CO2 has been developed by ILVO within the iSense project. Gas concentration measurements are taken by two tunable diode laser absorption spectroscopy (TDLAs) analysers (Axetris AG, Switzerland). Flow rate and inline pressure are also measured to provide additional information and improving the reliability of the gas measurement. An industrial-grade Arduino-compatible microcontroller unit is used to read and process sensor data in real-time. Measurements can be either stored locally on a Micro-SD card or instantly transferred to a PC/server via Ethernet. The microcontroller unit also consists of an LED screen and three control buttons, allowing users to view system status and to modify measurement schemes with ease.

The developed gas measurement system on NH3 measurement was validated in a fattening pig barn, and the measurement accuracy of the system was comparable to FTIRs, CRDs, QCLAs and PAs analysers and an impinger method. The high performance and low-cost of the system makes it a promising alternative to the high-end instruments in livestock research.

ISense and PASgeRUND collaboration – Shaojie Zhuang – update December 2017

PASgeRUND is a collaborative project between ILVO and Inagro, Innovation support center and Hooibeekhoeve aiming to reduce gas emissions from dairy farming facilities in Flanders. The first step of the project is developing a practically applicable method for quantifying emissions at barn and floor level. With a shared interest in cost-effective gas measuring techniques, iSense-PLF is now assisting PASgeRUND on investigating different gas sensing systems and measurement strategies in naturally ventilated dairy barns.

EC-PLF conference – Shaojie Zhuang - update August 2017

Some laboratory and field test results of a commercially available ammonia analyzer (Axetris® F200-A) will be presented at the EC-PLF 2017 conference, where the analyzer accuracy and potential improvement points will be demonstrated and discussed. 

Field tests in dairy cattle barn – Shaojie Zhuang - update August 2017

The scope of ammonia (NH3) sensor evaluation were extended to dairy facilities. Two 24-hour experiments were conducted in a naturally ventilated dairy barn and the performance of two Axetris F200-A tunable diode laser (TDLA) analyzers were compared with reference methods, i.e. open-path tunable diode laser and wet chemical method. The first experiment was carried out in cooperation with EnviVice (Netherlands), and the second solely by researchers from ILVO. Despite of the low ammonia concentration in the dairy barn, the performance of the Axetris TDLA was promising. Longer term experiments with the analyzer in a similar test set-up will soon be started.

Multisampler 2.0 – Shaojie Zhuang - update August 2017

The assembly of the second version of a multisampler is in progress. The system consists of 24 sampling lines and a common bypass line. The valves are controlled by a set of Arduino compatible microcontroller boards in a master-slave configuration. Sampling schemes can be viewed and changed directly via the on-board displays and buttons. The master controllers also processes and logs analog input signals of up to 18-bit resolution from four additional digital instruments. Upon completion the multisampler will be tested in a rearing facility for fattening pigs.

Field tests in fattening pig and beef cattle barn - Shaojie Zhuang – update May 2017

Multiple short-term field experiments were conducted in a fattening pig barn and a beef cattle barn to compare an Axetris F200-A tunable diode laser (TDLA) analyzer with reference methods in ammonia (NH3) measurement, i.e. Fourier Transform Infrared (FTIR) Spectroscopy and Cavity Ring-Down (CRD) Spectroscopy. In general the Axetris TDLA showed satisfying agreement with both of the reference instruments. Several influencing factors that may potentially affect the behavior of the Axetris TDLA were noticed and would be further investigated.

A major mistake in the FTIR measurement set-up was exposed and immediately fixed during the experiment. Impact of the mistake on the validity of the measurement were thereafter investigated. Based on the conclusion, some unexpected FTIR measurements from other experiments prior to the fix were well explained.

Multipoint gas measurement - Shaojie Zhuang - update Febr 2017

A prototype of a multichannel gas monitoring system was constructed and validated for measuring ammonia (NH3) emission from slurry in an experimental condition. High accuracy gas concertation measurement is performed by a tunable diode laser analyzer (TDLA). Inline pressure and flowrate in the sampling system are passively regulated using control valves, and channel switches are controlled by a programmable logic controller (PLC).

Construction of a second prototype will soon start including two major upgrades. Firstly the architecture design and the corresponding passive flow regulation will be improved to satisfy more diverse sampling requirement and operation conditions. Secondly the PLC will be substituted with an industrial-grade Arduino-compatible controller for better cost-effectiveness.

Indoor positioning system - Shaojie Zhuang- update Febr 2017

Positioning algorithms and geometry of base stations are two important factors influencing the performance of a range-based positioning system. A positioning algorithm has been developed to achieve optimal 2D positioning accuracy and computation efficiency in a barn for group-housed gestation sows. From the positioning data a trajectory of a moving object, such as a sow or a person,  can be recovered using a noise-reducing filter. Most of the previous tests intentionally avoided factors that might affect the performance of the system, such low battery and non-line-of-sight. Further tests will address the actual effect of those factors and feasibility of maintain satisfying positioning accuracy.

Figure: Localisation test with a sow