Computer vision tends the flocks (and herds)

We’ve written previously about computer vision in arable farming, the rise of precision agriculture and even its role in fish farming. Imaging crops and monitoring soil requires vision on a huge scale, but the technology has the advantage of capturing fairly static subjects. A drone can image an entire field with incredibly accurate results as a cabbage isn’t going to move to the other side of the field during the process. But what about monitoring the moving targets presented by livestock? That’s where the concept of Precision Livestock Farming (PLF) comes in, applying technology to ensure the optimum health and output of each individual animal.

The reasons for keeping an eye on your herd include checking the well being of animals, understanding when they’re ready for breeding and knowing it’s the right time to send them to market. RFID (Radio Frequency Identification) or GPS tags and collars have been around for a while, helping farmers keep track of their animals, but computer vision and AI are adding valuable and actionable information to such data. PLF combines data from sensors on or around the animals, often combined with environmental data such as temperature, ventilation or availability of water to guide farmers in making the best decisions to manage their livestock and output.

Precision farming in the cattle shed

Irish technology startup Cainthus offers a computer vision-enabled AI system for dairy farmers, monitoring cows and sending alerts and analysis via an app. Their ALUS Nutrition product uses smart cameras to watch over feeding activities to ensure that the correct amount of feed is accessible at the right time and refilled when needed. To differentiate between animals and feed, the system uses pixel pattern recognition to identify fodder and ensure cow comfort.

Icelandic farmers Aðalsteinn and Gardar Hallgrímsson have taken automation to the extreme with their 300-strong herd in Akureyri. In 2007, they invested nearly $1.5mn on feeding and milking systems and cleaning robots to improve the welfare of their cattle. Cows choose if and when to be milked, whether to be inside or out, and how much to eat. Sensors constantly monitor the cows and their milk output, and milk is tested to check vitamin levels. The animals have access to foam mattresses and a massage machine, and in 2011 the family opened a café on the site, using produce from the farm to cater to all the visitors who came to see the bovine paradise! What about ROI? Within a year, the farmers claim that milk produce was up 30% and vets bills down by 75%. Read the full story.

As happy as a pig in… a pen

Another real-time remote monitoring system, PigGuard, has been developed by Serket to watch over pigs. The system employs surveillance cameras and Deep Learning algorithms to apply anomaly detection to the herd, checking for and alerting farmers to abnormal behavior. Collected data translates into information about a pig’s physical activity, aggressiveness and feeding patterns to support swine well being.

Bristol Robotics Lab and Scotland’s Rural College have collaborated on a program to check for contented pigs. To train a vision system to understand what a happy pig looks like, researchers used 2D and 3D facial imaging to monitor pigs suffering from lameness compared to comfortable pigs. They also monitored animals that were given extra feeds to have examples of extra happy pigs. The results could help support animal welfare, and, the team suggest, could be a method for food standards agencies to enforce ethical farming. As a point of interest, in a separate research paper, “3D Vision for Precision Dairy Farming”[1], Niall O’Mahony et al. describe how Time of Flight (ToF) cameras are best suited to 3D imaging of livestock because they provide their own light source and can offer millimeter-level accuracy in real-time.

Focus on flocks

Octopus Robots have developed Poultry Safe and Scarifier to support the well being of barn-reared chickens. The two attachments fitted to the Octopus autonomous robot can sanitize the chicken shed and aerate litter on the floor to help reduce the risk of infection and injury, while navigating gently around the birds.

Putting computer vision and AI to use, Flox is another software provider offering multi-camera and night-vision monitoring inside hen houses to watch over clustering, smothering and bird weights.

The technology behind it

For a more in-depth understanding of which image processing techniques are best suited to monitoring the barnyard, a really good read is Quantum’s article in The Startup, demonstrating how to monitor pigs’ weights using instance segmentation with color and object detection methods.

Different camera technologies bring different benefits to remote monitoring; conventional image processing, thermal imaging and hyperspectral imaging all have a role to play in tending the flocks. The key is in how images are captured, processed and analyzed using Deep Learning and AI technologies to provide usable and informative data. Active Silicon are experts in computer vision and our range of cameras and interface boards are ideally suited to remote surveillance and video transmission. View our products and get in touch to see how our technologies could support your application whether its agricultural, industrial or medical.

[1] https://www.sciencedirect.com/science/article/pii/S2405896319324747