As an important link in the processing of jujube products, the qualities classification of jujubes have an important impact on improving the value of commodities. In this study, jujube target was extracted based on the RGB color space characteristics and then put into a black background through a mask. The data augmentation method combined deep convolutional generative adversarial networks and rigid transformation (RT) was used to improve the data richness of defective jujubes, effectively solve the imbalance problem between different types of jujube data. A composite convolutional neural network (CNN) method based on residual networks was designed to effectively solve the problem of misjudgment between jujubes with subtle defects and healthy jujubes. The overall results illustrated that the defect detection accuracy of the proposed scheme was 99.2%, which was superior to the widely used support vector machine and CNN methods. This work could be applied to the actual processing site and greatly improved the quality classification effect of jujubes.Practical ApplicationsCracks, peeling, wrinkles, and other defects have seriously affected the quality and value of jujubes, and the quality classification of jujubes is imperative. This paper proposes a set of deep learning schemes from three aspects of improving data quality, enhancing data richness, and designing more accurate and effective classification models. Experimental results show that this scheme can significantly improve the accuracy of jujube quality grading.
In production, due to natural conditions or process peculiarities, a single product often may exhibit more than one type of defect. The accurate identification of all defects has an important guiding significance and practical value to improve the planting and production processes. Concerning the surface defect classification task, convolutional neural networks can be implemented as a powerful instrument. However, a typical convolutional neural network tends to consider an image as an inseparable entity and a single instance when extracting features; moreover, it may overlook semantic correlations between different labels. To address these limitations, in the present paper, we proposed a feature-wise attention-based relation network (FAR-Net) for multilabel jujube defect classification. The network included four different modules designed for (1) image feature extraction, (2) label-wise feature aggregation, (3) feature activation and deactivation, and (4) correlation learning among labels. To evaluate the proposed method, a unique multilabel jujube defect dataset was constructed as a benchmark for the multilabel classification task of the jujube defect images. The results of experiments show that owing to the relation learning mechanism, the average precision of the three main composite defects in the dataset increases by 5.77%, 4.07%, and 3.50%, respectively, compared to the backbone of our network, namely Inception v3, which indicated that the proposed FAR-Net effectively facilitated the learning of correlation between labels and eventually, improved the multilabel classification accuracy.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.