Sorting and Identification Method of Camellia Seeds Based on Deep Learning

The sorting and identification of camellia seeds is a key technical link in the production and processing of camellia oil. The accurate removal of moldy camellia seeds can reduce the acidity of camellia oil, and the removal of camellia husks can improve the quality of subsequent production of camellia oil products. The traditional image processing methods face the problems of poor human selection features and complex feature extraction process. In this paper, a method for sorting and identifying camellia seeds based on deep learning is established. Based on the Resnet-18 network model, a transfer learning method is used to establish the camellia seed sorting and identification model. This Resnet-18 convolutional neural network proposes a residual function and Shortcut Connections to solve the problem of result degradation and the disappearance of gradient as the number of network layers deepens question. This model can independently recognize the effective features of the object, avoiding the complex process of artificially extracting features in traditional recognition algorithms. In addition, an image acquisition device is designed for the camellia seed mixture after the dehulling process and 1,200 sample pictures were collected. After testing, when the learning rate is 0.001 and the MiniBatchSize is 16, the recognition accuracy rate is 96.21%. The experimental data show that this method can effectively sort and identify Camellia seeds. This research provides a certain theoretical reference for the design of camellia seed sorting machinery.