Kuo Pang

Crack detection is a critical task in infrastructure monitoring, as surface-level deterioration can compromise the safety and operational integrity of structural systems. Manual inspection is labor-intensive, time-consuming, and impractical for large-scale deployment, reinforcing the need for automated visual inspection systems. However, deploying such systems in real time on edge devices remains challenging due to limited energy capacity and low-resolution image acquisition associated with embedded platforms. This paper proposes EECD-Net, an energy-efficient, multi-stage crack detection framework designed to improve both detection accuracy and energy efficiency. EECD-Net integrates three core modules: a Super-Resolution Convolutional Neural Network (SRCNN) to enhance image resolution while preserving structural details; a Spike Convolution Unit (SCU) based on Continuous Integrate-and-Fire (CIF) neurons to encode enhanced images into sparse spike sequences for low-power computation; and a Gated Attention Transformer (GAT) to adaptively fuse multi-scale features, capturing both long-range dependencies and fine-grained crack patterns. Experimental results on the CrackVision12K dataset demonstrate that EECD-Net achieves state-of-the-art crack detection performance, with 98.6% accuracy and an F1 score of 82.0%, while reducing energy consumption to 5.6 mJ per image. It outperforms existing methods, including DeepCrack2, HrSegNet, and Hybrid-Segmentor, in both accuracy and energy efficiency. These results suggest that EECD-Net offers a practical and effective solution for real-time, resource-constrained crack detection in intelligent infrastructure systems.

Online action detection and anticipation aim to understand current or upcoming actions in video streams. In industry, current artificial neural network (ANN)-based methods suffer from prohibitive energy consumption, fundamentally limiting their deployment on resource-constrained industrial devices. To bridge the gap between theory and application practice of informatics in industrial environments, we propose a novel knowledge distillation-based spiking neural network (KDSNN), which synergistically integrates bioinspired spike-driven processing with knowledge distillation, significantly reducing the energy consumption. Specifically, KDSNN includes a pioneering spiking neural network (SNN) architecture for online action detection and anticipation, which combines well-designed hierarchical spike convolutional neural network (CNN) block and spike Transformer block to capture spike-driven information. To further improve the performance of our SNN while maintaining low energy consumption, we introduce the knowledge distillation paradigm, which aims to utilize an expert-level ANN as a teacher to guide our SNN. Based on this, we propose a novel distillation loss, which consists of feature distillation and logit distillation. Notably, to address the cross-domain feature alignment in feature distillation, the optimal transport theory is employed to realize cross-domain knowledge transfer for the first time by minimizing the Wasserstein distance between continuous features (ANNs) and discrete features (SNNs). Through extensive evaluations on THUMOS14 and EPIC-Kitchen-100 datasets, the energy consumption of our KDSNN is only 27.1% and 10.0% of the state-of-the-art ANN-based method MAT. Equally importantly, the parameter count of our KDSNN is only 37.0% and 27.7% of MAT on THUMOS14 and EPIC-Kitchen-100, respectively.

In a world rich with linguistic-valued data, traditional methods often lead to significant information loss when converting such data into other formats. This paper presents a novel approach for constructing an interval linguistic-valued intuitionistic fuzzy concept lattice, which adeptly manages qualitative linguistic information by leveraging the strengths of interval-valued intuitionistic fuzzy sets to represent both fuzziness and uncertainty. First, the interval linguistic-valued intuitionistic fuzzy concept lattice is constructed by integrating interval intuitionistic fuzzy sets, capturing the bidirectional fuzzy linguistic information between objects, which encompasses both positive and negative aspects. Second, by analyzing the expectations of concept extent relative to intent, and considering both the membership and non-membership perspectives of linguistic expressions, we focus on the extraction of linguistic association rules. Finally, comparative analyses and examples demonstrate the effectiveness of the proposed approach, showcasing its potential to advance the management of linguistic data in various domains.