Invented by LI; Yan, VAN ADRICHEM; Paulus Jacobus Maria, ANIRAJ; Ananthu, VAN VLIET; Sjoerd Timo, SOZEN; Cagla, VAN DER SLUIS; Joram Robin, LELY PATENT N.V.
Modern animal husbandry is all about keeping animals healthy and comfortable. But tracking animals as they move around a barn or field is tough. This new patent application introduces a smart camera system that uses images from above to accurately monitor and track each animal, even when they walk between different camera views. Let’s break down the reasons for this invention, the science behind it, and what makes it truly special.
Background and Market Context
Farmers and ranchers want healthy animals, happy animals, and efficient farms. Today, many farms use technology to monitor animals. They want to know if a cow is eating, lying down, or walking. These details help spot health problems early and improve care. But when lots of animals move around in a big open space, keeping track of each one is really hard.
In barns and open fields, animals roam freely. They often walk from one spot to another, and sometimes several animals gather close together. Cameras can help, but a single camera can’t see everywhere. Farms usually need several cameras, each looking at a part of the area. The main problem? If a cow moves from one camera’s view to another, the system must recognize it’s the same cow and keep tracking it. If the system gets it wrong, important data about that animal is lost or mixed up.
Older systems often use tags, collars, or other gadgets on each animal. These can fall off, break, or bother the animals. Some new systems use image tracking, but even these can mix up animals when they cross camera borders. This means farmers may get the wrong information about which cow is eating, which is lying down, or which is sick or healthy.
What the market really needs is a system that can:
– Watch many animals at once, without needing tags or collars.
– Use several cameras to cover the whole barn or field.
– Know when an animal moves from one camera’s area to another, and keep tracking the right animal.
– Tell if an animal is standing, lying down, eating, or moving.
With better tracking, farmers can see changes in animal behavior right away. This helps find sick animals faster, keep animals comfortable, and make farms more efficient. The industry is looking for simple, non-intrusive, and reliable solutions.
Scientific Rationale and Prior Art
Let’s look at how earlier technologies have tried to solve this problem and where they fall short.
Older animal monitoring systems often use cameras high up in the barn, pointing down. These cameras watch for movement and sometimes can spot if an animal is lying down or standing. Some systems use special tags or collars on each animal. These tags send signals to a computer, which tracks where each animal is. But, tags can get lost, break, or bother the animals. Some animals even figure out how to remove or chew on them.
Some more advanced systems try to use just the camera images, without tags. These image-based systems use computer programs to find the shapes of animals in the pictures. They use smart software, like neural networks, to figure out if each shape is a cow, a goat, or another animal. But these systems still have trouble when animals move from one camera’s view to another. If two cows are close together, or if a cow walks out of one camera and into another, the computer gets confused. It might think the cow is a new animal, or mix up which cow is which.
Some patents show ways to stitch together images from several cameras into one big picture. They try to match the shapes of animals seen in two different cameras. But, this is tricky because cameras are often at different angles. If a cow is standing up, it looks taller in the picture. If it’s lying down, it looks flatter and takes up less space. The computer must figure out which shape in one camera matches the shape in another camera, even if the cow’s pose is different. If the system makes a mistake, it may count the same cow twice or lose track of a cow altogether.
Past inventions have tried to fix these problems in different ways. Some use special lighting or try to make sure the cameras are perfectly aligned. Others use smart algorithms to guess which animal is which based on where they are standing. Some even use face recognition, but this is very hard with animals, especially when they turn away or group together.
The main issues with previous systems include:
– Confusing animals when they move between cameras.
– Trouble telling animals apart when they are close together.
– Difficulty matching the shape of a standing animal with the same animal when it lies down.
– Needing extra gadgets or tags on the animals.
The new patent builds on these ideas, but adds a clever twist. It uses a special trick to project the animal’s shape from where it is detected (like the middle of a cow’s body) down to the floor of the barn. This helps the computer match the same animal across different camera views, even if the animal is standing in one picture and lying down in another. By knowing if the animal is standing or lying, the computer can use the right kind of projection and make a better match.
Invention Description and Key Innovations
Now, let’s look at how this new system actually works and what makes it stand out.
Imagine a big barn with cows walking around. Several cameras are placed above, watching the animals from different spots. Each camera takes pictures or videos of its own part of the barn. Some areas may be seen by more than one camera at the same time. This is good, because it means there is overlap, and the system can use more than one angle to look at each animal.
The computer connected to the cameras does several things:
First, it finds the shape or “silhouette” of each animal in each picture. It then draws a box around each animal, called a “bounding box.” The computer then uses smart image analysis, like neural networks, to figure out if the animal inside each box is standing or lying down. This is important, because a standing cow is higher up from the floor, while a lying cow is closer to the ground.
Then comes the clever part. The system projects each animal’s box from its height (like the middle of its body) down to the floor of the barn. If the cow is standing, the projection is longer, because the cow is higher up. If the cow is lying down, the projection is shorter. By doing this for all animals in all pictures, the system makes sure each animal’s position is shown on the same flat “plane”—the floor.
Next, the system stitches together the images from all the cameras, using the floor as the common reference. It lines up the projected boxes from different cameras. If two boxes from different cameras overlap on the floor, the computer knows they are showing the same animal. To make this matching process very accurate, the system can use a special algorithm called “bipartite matching.” This math trick helps pair up boxes from different cameras, even if some animals are only seen by one camera.
The system can spot if a cow is standing near the feed area, lying in a cubicle, or moving from one spot to another. Because it knows which animal is which, even when they walk from one camera’s view to another, it keeps a record of each animal’s actions. Farmers can then see which cows are eating, which are resting, and which may need care.
If the system did not use this special projection step, it might mix up animals when they cross camera borders. For example, two cows standing side by side may look very similar from above, and if one lies down, its shape changes. By projecting each animal’s position down to the same level, the system avoids these mistakes and keeps better track of each animal.
Another smart feature is that the system can use different ways to tell if a cow is standing or lying down. It can use neural networks trained on lots of animal pictures. It can also look at where the animal is—in the resting area, it’s more likely to be lying down. Or it can look at the direction the animal is facing. All these clues help the computer make the right choice.
This makes the whole system more reliable, with no need for tags, and with fewer errors. The data collected is better, which means farmers can make better decisions. They know right away if an animal changes its routine, which can be an early sign of illness or stress.
Conclusion
This new animal husbandry system patent brings a fresh and practical solution to a tough problem in modern farming. By combining smart cameras, simple projections, and clever computer algorithms, it keeps track of each animal as it moves freely around a barn or field. It works without tags, and can tell when animals stand up, lie down, or move between camera views. This means better data, healthier animals, and more peace of mind for farmers. As farms grow and technology becomes more important, solutions like this will shape the future of animal care.
Click here https://ppubs.uspto.gov/pubwebapp/ and search 20250228208.