Intelligent poultry seedling sorting system based on convolutional neural network

Abstract

In view of the outstanding labor costs and high health and safety risks in the traditional poultry seedling sorting process, an automatic equipment suitable for the sorting of poultry seedlings was designed through the research on the extraction of the characteristics of a variety of poultry species. According to the different types of poultry seedlings, there are many different sorting requirements for chicken seedlings, duck seedlings, goose seedlings, etc., thus providing a classification and recognition method that combines visual information technology and convolutional neural network (CNN) technology to form a hardware and software collaboration framework to meet the functional requirements for image acquisition, processing and real-time detection of different poultry seedlings. First of all, we use camera technology to extract the features on the shape of the bird seedlings into the system and wait for the next step of correlation analysis; Then, based on CNN technology, different species of the same bird vaccine are classified; Finally, sorting is realized through hardware collaborative processing. For each kind of poultry seedlings, multiple trainings shall be conducted, and the corresponding database shall be established to improve the sorting accuracy. The results showed that the recognition accuracy of different types of poultry seedlings was 100%, the recognition accuracy of different types of poultry seedlings of the same species was 97.7%, and the recognition speed was 0.049 seconds, which could meet the requirements of efficient sorting of 8000 poultry seedlings per hour.