Dong Ren∗ and Bin Li∗∗



category attribute features from multi-scale recurrence plot on the basis of synchrosqueezed wavelet transform (SWT) and construct a two-level feature extraction method for multi-class EEG. The average classification accuracy shows that the proposed method has good effectiveness and robustness. Deep learning as an important branch of AI has received extensive attention in the field of remote sensing. Such as classifying the vegetation into subclasses is significant for some applications such as ecological protection and vegetation mapping. In “Vegetation Classification by Multi-scale Hierarchical Segmentation on GF-2 Remote Sensing Image [3]”, Wugu et al. propose a new method for classifying vegetation into subclasses using GF-2 remote sensing image on the basis of multi-scale segmentation. The GF-2 remote sensing image is separated into objects in different hierarchical levels by using image features such as spectral, shape, texture and others. Experiment is performed with the image along the Yangtze River in the Dianjun District of Yichang City and shows the efficiency of the method. In addition, the application of deep learning in the field of target detection also shows its powerful function. In 1 “Research on the Application of RetinaNet Combined with Adaptive Learning Rate Attenuation Strategy in Vehicle Type Detection [4]”, Xu et al. address the issues of accuracy and speed of the current target detection algorithm, by using deep learning model. They use RetinaNet as the basic framework for vehicle type detection and propose an adaptive learning rate attenuation (ALRA) on the basis of least squares. The experimental results demonstrated ALRA can improve model convergence effectively. Non-destructive testing of heavy metals is an application of AI in environmental automation systems. In “Application of Backpropagation Neural Network for Soil Heavy Metal Modelling [5]”, Li et al. build accurate models for rapid detection of heavy metal pollution in soil with X-ray fluorescence (XRF) spectrometer. A novel error back-propagation artificial neural network learning (BPANN) algorithm optimized by Levenberg–Marquardt (LM) algorithm is selected to establish accurate models for quantitative detection of soil heavy metal XRF. Experimental results demonstrated the promising performance of BPANN-LM. To improve the accuracy of heavy metal detection, in “An Improved Boosting-BiPLS Models Based on Weight Adjustment for Soil Heavy Metal Content Prediction [6]”, Ren et al. propose an improved Boosting-BiPLS model to detect the heavy metal content in soil. From bias oriented, the weights of samples are adjusted on the basis of the relative deviation of the samples and the weights of base models are dynamically calculated by the spectral similarity. The results show the improved Booting-BiPLS model is more precise and stable than previous models. Non-destructive rapid detection of apples is an another application of AI in environmental automation systems. In “A Proprietarily Developed Bionic Olfactory System Used for Rapid Detection of Deteriorated Refrigerated-Stored Apples [7]”, Tian et al. design a high-sensitivity, lowcost portable electronic nose detection system, using rapid detection and early-warning of apple deterioration. And a non-destructive detection model for refrigerated-stored apples is constructed on the basis of PCA/KPCA and BPNN/SVM. Acknowledgement We hope that the special issue will help researcher’s present novel contributions in a related field in the future. And we thank Prof. Simon X. Yang, the Editor-in-Chief of the International Journal of Robotics and Automation, for his support during the SI production process.

Important Links:

Go Back