Xueda Huang

Driver distraction detection not only effectively prevents traffic accidents but also promotes the development of intelligent transportation systems. In recent years, thanks to the powerful feature learning capabilities of deep learning algorithms, driver distraction detection methods based on deep learning have increased significantly. However, for resource-constrained onboard devices, real-time lightweight models are crucial. Most existing methods tend to focus solely on lightweight model design, neglecting the loss in detection performance for small targets. To achieve a balance between detection accuracy and network lightweighting, this paper proposes a driver distraction detection method that combines enhancement and global saliency optimization. The method mainly consists of three modules: context fusion enhancement module (CFEM), channel optimization feedback module (COFM), and channel saliency distillation module (CSDM). In the CFEM module, one-dimensional convolution is used to capture information between distant pixels, and an injection mechanism is adopted to further integrate high-level semantic information with low-level detail information, enhancing feature fusion capabilities. The COFM module incorporates a feedback mechanism to consider the impact of inter-layer and intra-layer channel relationships on model compression performance, achieving joint pruning of global channels. The CSDM module guides the student network to learn the salient feature information from the teacher network, effectively balancing the model’s real-time performance and accuracy. Experimental results show that this method outperforms the state-of-the-art methods in driver distraction detection tasks, demonstrating good performance and potential application prospects.

Distracted behavior detection is an important task in computer-assisted driving. Although deep learning has made significant progress in this area, it is still difficult to meet the requirements of the real-time analysis and processing of massive data by relying solely on local computing power. To overcome these problems, this paper proposes a driving distraction detection method based on cloud–fog computing architecture, which introduces scalable modules and a model-driven optimization based on greedy pruning. Specifically, the proposed method makes full use of cloud–fog computing to process complex driving scene data, solves the problem of local computing resource limitations, and achieves the goal of detecting distracted driving behavior in real time. In terms of feature extraction, scalable modules are used to adapt to different levels of feature extraction to effectively capture the diversity of driving behaviors. Additionally, in order to improve the performance of the model, a model-driven optimization method based on greedy pruning is introduced to optimize the model structure to obtain a lighter and more efficient model. Through verification experiments on multiple driving scene datasets such as LDDB and Statefarm, the effectiveness of the proposed driving distraction detection method is proved.

Agile Product Line Engineering (APLE), a relatively new approach combining the two successful methods of Agile Software Development (ASD) and Software Product Lines (SPLs), makes product lines more responsive to ever-changing customer needs or market changes. However, SPLs often fail to keep up with market demand due to high coordination costs, slow development processes, and long release cycles in the case of frequent changes in business requirements; in agile software projects, the lack of a unified specification for describing requirements leads to high coordination costs and inconvenient requirement management. Some studies in the literature have proposed optimized approaches to integrate ASD and SPLs, but they still have not covered all aspects of APLE’s characteristics, and software resource reuse is rarely considered in these approaches during product line development. In view of this, we propose a collaborative framework of agile product line engineering for software resource reuse, namely ScrumOntoSPL. The ScrumOntoSPL approach efficiently merges ASD and SPL based on the agile method Scrum, SPL architecture, and ontology technology. In ScrumOntoSPL, uniform requirement specification is constructed by utilizing ontology, and the Matching Requirement with Component (MRC) process is designed to match product new requirements and software resources stored in a resource pool. In addition, we evaluated the proposed framework and approach with CMMI. In the end, a case study of a software development tool called IMC-Tool based on ScrumOntoSPL for a universal Instrument Microcontroller Chip (IMC) is discussed. The IMC-Tool case illustrates that the ScrumOntoSPL has the advantages of dynamically managing demand changes, enhancing software resource reuse, reducing coordination costs, and reducing time to market.

Safety helmet wearing plays a major role in protecting the safety of workers in industry and construction, so a real-time helmet wearing detection technology is very necessary. This paper proposes an improved YOLOv4 algorithm to achieve real-time and efficient safety helmet wearing detection. The improved YOLOv4 algorithm adopts a lightweight network PP-LCNet as the backbone network and uses deepwise separable convolution to decrease the model parameters. Besides, the coordinate attention mechanism module is embedded in the three output feature layers of the backbone network to enhance the feature information, and an improved feature fusion structure is designed to fuse the target information. In terms of the loss function, we use a new SIoU loss function that fuses directional information to increase detection precision. The experimental findings demonstrate that the improved YOLOv4 algorithm achieves an accuracy of 92.98%, a model size of 41.88 M, and a detection speed of 43.23 pictures/s. Compared with the original YOLOv4, the accuracy increases by 0.52%, the model size decreases by about 83%, and the detection speed increases by 88%. Compared with other existing methods, it performs better in terms of precision and speed.