In astronomy, it is important to categorize celestial bodies by classifying collected spectral data. The currently available methods present unsatisfactory spectral classification accuracy and incur high computing costs. We propose a celestial spectral classification network based on a residual and attention based convolutional network (RAC-Net). In this network, convolution operations can extract shallow and deep features of spectral data and classify them without relying on redshifts. The residual mechanism can augment the depth of the network and make training more efficient. The attention mechanism allows the network to focus on specific bands and specific features, rendering the learning more targeted. To evaluate the performance of the RAC-Net, we conducted a comparative test using a celestial spectral data set that consisted of 70,000 spectra collected by the large sky area multi-object fiber spectroscopic telescope. The experimental results showed that the classification accuracy of our network was up to 98.92%. Compared with the leading one-dimensional, convolutional neural network 1D SSCNN model, the RAC-Net presented higher classification accuracy and fewer network parameters.

Applying machine learning to lemon defect recognition can improve the efficiency of lemon quality detection. This paper proposes a deep learning-based classification method with visual feature extraction and transfer learning to recognize defect lemons (i.e., green and mold defects). First, the data enhancement and brightness compensation techniques are used for data prepossessing. The visual feature extraction is used to quantify the defects and determine the feature variables as the bandit basis for classification. Then we construct a convolutional neural network with an embedded Visual Geometry Group 16 based (VGG16-based) network using transfer learning. The proposed model is compared with many benchmark models such as K-nearest Neighbor (KNN) and Support Vector Machine (SVM). Results show that the proposed model achieves the highest accuracy (95.44%) in the testing data set. The research provides a new solution for lemon defect recognition.

No abstract