Design And Implementation Of Holstein Cow Image Recognition Management System Based On SIFT

With the rapid development of science and technology,the requirements for safety and high quality of dairy products are steady raised.Therefore,it is an urgent task to optimize the modern and intelligent management system of dairy farms.Cow identification is an important part of the management system of dairy cattle.In the past,the dairy farms were mainly identified by ear tags installed on the ears of cows.The efficiency of the identification was low and it was easily affected by human factors.At present,some dairy farms have begun to use electronic tags,but this method has higher cost and is difficult to be popularized on a large scale.Both the installation of ear tags or electronic Labels can cause physical and mental damage to the cows and have a certain impact on their behavior.Using image recognition technology to identify individual dairy cattle with her biological features shows strong stability and individual differences.The non-contact,high precision and low cost individual recognition methods based on image processing are more and more popular recently.Cow's recognition is based on the unique markings on the body of the cow,which is lifetime invariance.In this thesis,author has completed the following tasks:1.Dairy cattle image acquisition.In order to study the recognition accuracy of cow images,the image acquisition of dairy farms was carried out.In the process of collection,taking into account the complexity of the environment,and in order to not disturb the work of dairy farms,we design a special solution to capture video data with uniform background when the dairy cattle walking through the narrow aisle.2.Technical scheme.A new scheme of scale invariant feature transformation algorithm based on L component of L?? color space is proposed by comparing and studying the transformation of many color spaces.The SIFT algorithm is applied to the cow recognition to use local features in scale space.Through experiments being carried out in a variety of color space,L?? color space is used as the recognition space owning rotation,zoom and brightness invariant,as well as feature descriptor with changes of view,mode transform and noise with good stability.Experimental results show that the proposed scheme can recognize cow images effectively,and the recognition accuracy can be morethan 98%.3.System design and implementation.On the basis of theoretical research,this paper develops a Holstein cow image recognition management system based on windows and OpenCV platform.The system uses MFC as a framework for development,using the Visual Studio 2010 as development tools,and SQL Server 2008 as the database.The system mainly includes user registration module,data acquisition module,cow identification module,information browsing module,cow information generation module and so on.The system is well to convert the related algorithms to the interface of visual operation.The test shows that the system can be more convenient and practical.