Digital imaging weighing of live animal with 2D Camera and artificial database
| Reference number | |
| Coordinator | Smart Agritech Solution of Sweden AB |
| Funding from Vinnova | SEK 297 134 |
| Project duration | May 2018 - December 2018 |
| Status | Completed |
| Venture | Innovative Startups |
Purpose and goal
The goal of testing cameras and images of live animals in self-taught image recognition is reached.
Expected results and effects
The result of the image processing was successful. It is possible to distinguish the animal in the image, regardless of which camera was used. The images can be compared and interpreted with image recognition algorithms. By teaching the system that a certain appearance (of an animal) corresponds to a certain weight, it can be taught to weigh an animal from a digital image.
Planned approach and implementation
The arrangement of collecting images of live animals and driving them through an image recognition system proved successful. The system, regardless of camera, can analyze the size and thus the weight of the animal. Even though reliability does not reach 100%, the fact that pictures are taken quite often means a fully useful growth curve is provided. Pigs are weighed with an accuracy of + -1.5kg. Cattle have a slightly higher discrepancy due to their dissymmetry + -5% which still is more than enough.