In the past, pig farmers could only carry out extensive management in the process of breeding. In fact, each pig's eating preference, food intake and health status are different. To achieve personalized management, we must first identify them. With pig face recognition technology, these problems can be solved easily. The media saw that the operator could get the pig's serial number, parents, strain and other information just by raising his mobile phone and scanning it at a pig.
By yesterday morning, 16882 1 a big pig from 1692 pig farm had participated in the pig face recognition experience. Dr. Chen Yaosheng, the chief scientist of the national pig industry technology system and a professor at Sun Yat-sen University, admitted that the pig face recognition technology still faces many problems. "People will take pictures in front of the camera when they fly, but pigs will not." He said that the pig's cooperation was low and he was a little disobedient when taking pictures. The media also saw that the input of pig face information needs multi-angle scanning, and pigs are either lying down or squinting, with more patterns than when humans take selfies. It takes the operator a lot of time to finish taking pictures. One of the biggest differences from face recognition is that faces are highly differentiated, while pigs are multiple viviparous animals and look alike. In addition, pigs will be affected by the living environment and fight with each other, making their faces dirty and difficult to recognize. In the experimental field of pig face recognition technology, the recognition rate of sows is 98%, and that of pigs is 85%. In the field of testing and experience, this ratio is even lower. However, Chen Yaosheng pointed out that they can greatly improve the technology by adjusting the model and algorithm.
According to reports, one of the application scenarios of pig face recognition technology is the fine individual management of pigs. Chen Yaosheng, for example, said that when pigs enter the channel to eat, the system can recognize them by "brushing their faces" and put them into the feed according to their actual situation. In addition, if there is a problem with the pig, it can be marked with the help of auxiliary equipment so that the staff can follow up. From the perspective of food safety, slaughterhouses can identify which slaughterhouse each pig comes from, making monitoring more efficient and easy.
Wu, president of Guangdong Pig Industry Association, believes that pig face recognition will bring subversive changes to the industry and become the most critical information products such as identity recognition, food safety and credit construction. At the same time, this technology can lay a foundation for the establishment of industry big data, and quickly connect distributed databases through efficient identification, so that everyone can quickly understand the whole process of products, from seeing to believing, to establishing credit and creating new value chains.
Chen Yaosheng told the media that they will further improve the accuracy of identification, expand the scope of application of varieties and weights, and open up isolated islands such as breeding, nutrition, veterinary, environmental control and production management. "In the past, pigs lost weight and needed to find nutritionists and veterinarians to see a doctor. We can connect these links through efficient and accurate data interpretation, and the level of pig raising will be greatly improved. "