Deep Learning for Information Fusion and Pattern Recognition

There is a large amount of data from different types of sensors, for instance, multispectral electro-optical/infrared (EO/IR) and computed tomography/magnetic resonance (CT/MR) images, among others. How to take advantage of multimodal data for object detection and pattern recognition is an active field of research. Information fusion (IF) is used for enhancing the performance of pattern classification, while deep learning (DL) technologies, including convolutional neural networks (CNNs), are powerful tools for improving object detection, segmentation, and recognition. It is viable to combine DL and IF to boost the overall performance of pattern classification and target recognition. Such combinations of powerful techniques may exploit the deeply hidden features of the multimodal, spatial, or temporal data. This reprint presents cutting-edge research utilizing DL and IF techniques. Key research areas include image and video analysis, covering topics such as super-resolution, object detection, semantic segmentation, video captioning, and text processing, including labeling enhancement and screening misinformation. Biometric applications explore innovations in human identification using facial and finger vein recognition, facial micro-expression analysis, and fatigue detection. Advanced applications extend to handwritten recognition, tracking supermarket customer behavior, parcel sorting, predicting road surface conditions, and plastic waste classification.