Vision improves dairy farm milking practices

Share this on social media:

Vision sensors incorporating time-of-flight technology are being used to automate milking in dairy farms. The teat finding sensor, called Tracker 4000, offers advantages of improved yield performance and farm productivity, livestock well being, enhanced reliability, increased speed of farm operations and profitability, and improved product quality in the milking process. Tracker 4000 was developed by GEA Farm Technologies alongside LMI Technologies.

A key component in automated milking systems is the vision sensor (combination of LED lighting, lasers, cameras, real time image processing and geometric pattern recognition software) that identifies and locates the teat positions on the cow's udder as it steps into the milking box. The information derived from the sensor is used to guide a robot arm to attach the milking cups to the teats. The entire process of locating teats and attaching cups must be fast, accurate and safe, without disturbing or stressing the cow.

With cows being highly individual animals in behaviour and appearance, the challenges for the vision sensor to cope with varying colour and a live, moving animal in a naturally dirty environment are extremely complex, along with the mandatory high degree of accuracy that is required.

The Tracker 4000 uses 3D time-of-flight imaging technology to catch the entire scene in one simultaneous exposure. High speed image processing rapidly and simultaneously identifies both teat and milking cups locations in the sensor field of view and guides the milking cups to their respective teat.

Real-time image processing algorithms running in the LMI's FireSync processor located inside the sensor use proprietary software to extract teat and milking cup locations in the images, ignoring other objects in the field of view, like a cow's leg or a swinging tail. The final result, coordinate positions for teats and milking cup locations in a predefined coordinate system, is delivered to the robot controller via an Ethernet connection.

The high speed, low weight and small size of the vision sensor allows it to be mounted on the actuator end of the cup attachment robot arm and thereby enables a simultaneous verification of the milking cup position before attachment. The Tracker 4000 is a solid-state intelligent vision sensor, with no moving parts or laser light, combining new imaging technology advances with specialised real-time image processing and optics.