The existing deep learning-based Personal Protective Equipment (PPE) detectors can only detect limited types of PPE and their performance needs to be improved, particularly for their deployment on real construction sites. This paper introduces an approach to train and evaluate eight deep learning detectors, for real application purposes, based on You Only Look Once (YOLO) architectures for six classes, including helmets with four colours, person, and vest. Meanwhile, a dedicated high-quality dataset, CHV, consisting of 1330 images, is constructed by considering real construction site background, different gestures, varied angles and distances, and multi PPE classes. The comparison result among the eight models shows that YOLO v5x has the best mAP (86.55%), and YOLO v5s has the fastest speed (52 FPS) on GPU. The detection accuracy of helmet classes on blurred faces decreases by 7%, while there is no effect on other person and vest classes. And the proposed detectors trained on the CHV dataset have a superior performance compared to other deep learning approaches on the same datasets. The novel multiclass CHV dataset is open for public use.
Relevance feedback is a mechanism to interactively learn a user's query concept online. It has been extensively used to improve the performance of multimedia information retrieval. In this paper, we present a novel interactive pattern analysis method that reduces relevance feedback to a two-class classification problem and classifies multimedia objects as relevant or irrelevant. To perform interactive pattern analysis, we propose two online pattern classification methods, called interactive random forests (IRF) and adaptive random forests (ARF), that adapt a composite classifier known as random forests for relevance feedback. IRF improves the efficiency of regular random forests (RRF) with a novel two-level resampling technique called biased random sample reduction, while ARF boosts the performance of RRF with two adaptive learning techniques called dynamic feature extraction and adaptive sample selection. During interactive multimedia retrieval, both ARF and IRF run two to three times faster than RRF while achieving comparable precision and recall against the latter. Extensive experiments on a COREL image set (with 31,438 images) demonstrate that our methods (i.e., IRF and RRF) achieve at least a 20% improvement on average precision and recall over the state-of-the-art approaches.
To study the mechanism of vault lining under different void heights and verify the strengthening effect of the attached steel plate, a CDP (concrete-damaged plasticity) model and the XFEM (extended finite element method) were used to construct the local numerical model of the vault void, and an experiment was carried out for verification. The strengthened structure of the steel plate was assembled with a combination of a two-component epoxy adhesive and chemical anchor bolts. Five lining models with various void thicknesses, together with their strengthened models, were evaluated. The results of the established numerical model were compared with the experimental results in terms of failure mode, vertical displacement, and load-deformation results. The results of the two numerical models were in good agreement with the experimental results, revealing the failure mechanism of the vault lining. The rigidity of the specimen after steel plate strengthening was significantly improved. When the void height was one-fourth of the secondary lining thickness, the lining cracks were reduced from 14 to 4, and the distribution width of the cracks was also reduced from 1.047 to 0.091 m after steel plate strengthening. The level of damage caused by cracking was significantly reduced, which proves the effectiveness of the surface-sticking method for steel plate strengthening.
Classifying high-dimensional numerical data is a very challenging problem. In high dimensional feature spaces, the performance of supervised learning methods suffer from the curse of dimensionality, which degrades both classification accuracy and efficiency. To address this issue, we present an efficient feature selection method to facilitate classifying high-dimensional numerical data. Our method employs balanced information gain to measure the contribution of each feature (for data classification); and it calculates feature correlation with a novel extension of balanced information gain. By integrating feature contribution and correlation, our feature selection approach uses a forward sequential selection algorithm to select uncorrelated features with large balanced information gain. Extensive experiments have been carried out on image and gene microarray datasets to demonstrate the effectiveness and robustness of the presented method.
Loess tunnels often undergo large-scale deformation with complex spatial and temporal distribution. Mastering the characteristics of spatial and temporal deformation is conducive to precise policy implementation and the control of large deformation of the tunnel. In this study, relying on the Yulinzi Tunnel in Gansu Province, China, based on 3D laser scanning technology, the tunnel was monitored for a short period of 24 h and a long period of 36 days. The refined characteristics of the temporal and spatial deformation of the representative points of the interrupted surface, the tunnel face, and the excavation mileage during the excavation process of the three-bench and seven-step method of the tunnel were analyzed. The results show that the tunnel’s arch has large deformation, and there is twisting deformation. The distribution of the overall deformation of the tunnel is related to the excavation sequence, showing a stepped deformation law. With the construction of the following excavation process, the deformation rate of the tunnel always indicates the characteristics of significant in the early stage and small in the later stage, and the overall deformation changes with time in accordance with the distribution law of the exponential function. The research results provide a reference for predicting the deformation development trend of loess tunnels and providing reasonable deformation control methods.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.