Computer Vision News - September 2016

Every month, Computer Vision News reviews a successful project. Our main purpose is to show how diverse image processing applications can be and how the different techniques can help to solve technical challenges and physical difficulties. This month we review a software for Cow Feeding and Optimization of Milk Output , developed by RSIP Vision for one of its Precise Agriculture clients. Do you have a project in computer vision and image processing? Contact our consultants . Our client wanted a system to provide the optimal quantity of food to cows, in order to maximize the production of milk for each individual animal . This involves giving each cow the quantity of food which best fits it: a fat cow does not increase production by overfeeding it. A thin cow does not generally produce as much milk as it would, were it properly fed. Bringing these two cows into the “normal” group prevents food waste on one side and increases milk output on the other. A veterinary study found a correlation between the cow’s fat and the size of her rear part. This enables to estimate how fat the cow is by measuring the puffiness of its rear, which is done by technologies of image processing . At the beginning we wanted to photograph the part, then we understood that the simplest solution was a laser scanner : 3 times a day, the cow slowly walks its way through the milking process line. That is the ideal spot for a laser scanner, which scans red markers on the cow’s back at video frequency (25 frames per second). The process developed by RSIP Vision stitches together 10 or 20 images into one, in which the markers describe the extent of the cow’s back. This is compared to the veterinary’s model to assess if the cow is fat, thin or normal. With this data in hand, the veterinary decides the daily quantity of food intake which is proper for that cow. Project Cow Feeding and Optimization of Milk Output The challenge is not as easy as it sounds: lighting conditions in a cowshed are far from optimal and dirt is everywhere, which makes the problem very different from regular conveyor belt imaging. Using a laser scanner solved those problems : the devices scans the red markers and disregards all kinds of noise. The result proved extremely effective and allowed our client to sell his system in international markets: first, it is entirely automatic and does not require any intervention. Second, it is fast. Finally, being repeated 3 times per day, it gives a very quick feedback about the cow’s condition and enables immediate corrective measures about food intake. Computer Vision News Project 35