Bingqiang Huang

Dr. Bingqiang Huang


Research Focus & Metrics

Academic Career Length

6Years (2021-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
7 August 2022 in Sensors
Verified
Journal Article
Open access
PDF file
XXXinjian Xiang
Xinjian Xiang
KLKehan Li
Kehan Li
BHBingqiang Huang
Bingqiang Huang
YCYing Cao
Ying Cao
The essential factors of information-aware systems are heterogeneous multi-sensory devices. Because of the ambiguity and contradicting nature of multi-sensor data, a data-fusion method based on the cloud model and improved evidence theory is proposed. To complete the conversion from quantitative to qualitative data, the cloud model is employed to construct the basic probability assignment (BPA) function of the evidence corresponding to each data source. To address the issue that traditional evidence theory produces results that do not correspond to the facts when fusing conflicting evidence, the three measures of the Jousselme distance, cosine similarity, and the Jaccard coefficient are combined to measure the similarity of the evidence. The Hellinger distance of the interval is used to calculate the credibility of the evidence. The similarity and credibility are combined to improve the evidence, and the fusion is performed according to Dempster’s rule to finally obtain the results. The numerical example results show that the proposed improved evidence theory method has better convergence and focus, and the confidence in the correct proposition is up to 100%. Applying the proposed multi-sensor data-fusion method to early indoor fire detection, the method improves the accuracy by 0.9–6.4% and reduces the false alarm rate by 0.7–10.2% compared with traditional and other improved evidence theories, proving its validity and feasibility, which provides a certain reference value for multi-sensor information fusion.

12 June 2023 in IEEE Access
Verified
Journal Article
Open access
ZFZhengshun Fei
Zhengshun Fei
JGJunhao Guo
Junhao Guo
HGHaibo Gong
Haibo Gong
LYLubin Ye
Lubin Ye
EAEric Attahi
Eric Attahi
BHBingqiang Huang
Bingqiang Huang
Convolutional neural network (CNN) is quite popular in computer vision, especially in image classification with excellent performance. However, limited by the convolution kernels, CNN-based classifiers are hard to extract global feature from the original image, while exact object locations in the environment are included in the global feature. One popular way to improve global feature extraction performance is to use graph neural network (GNN) which can aggregate global information through the connection relationship of different nodes. In this work, a novel end-to-end graph neural network architecture is proposed, in which local and global-attention feature are used simultaneously to achieve more accurate predictions. In this architecture, a CNN block is designed to learn local feature and graph convolutional neural network (GCN) is used to learn global feature. Global-attention feature for final prediction is down-sampled from global feature by the proposed global multi-head self-attention pooling (GMSAPool) based on self-attention mechanism, which reconstructs the input graph by introducing virtual node and automatically assigns different weights to each node to obtain a more representative global-attention feature. In addition, the proposed architecture can be trained without converting images to graphs in advance, and the computational burden can also be reduced. This approach is demonstrated on three open datasets (Agricultural Disease, Caltech256 and CIFAR-100) to validate the effectiveness. The determined experimental results showed that: (1) The proposed model achieve 84.46%, 77.80%, and 83.33% on the Macro-F1 in three datasets respectively, improving over the best baselines; (2) Global-attention feature that is more conducive to the final prediction is extracted by GMSAPool from numerous nodes, in which Macro-P,Macro-R and Macro-F1 are respectively improved 3.655%,1.12%,2.715% on average.

13 October 2022 in Computational Intelligence and Neuroscience
Verified
Journal Article
Open access
PDF file
YCYinze Chu
Yinze Chu
XXXinjian Xiang
Xinjian Xiang
YWYilin Wang
Yilin Wang
BHBingqiang Huang
Bingqiang Huang
An improved Ghost-YOLOv5s detection algorithm is proposed in this paper to solve the problems of high computational load and undesirable recognition rate in the traditional detection methods of pavement diseases. Ghost modules and C3Ghost are introduced into the YOLOv5s network to reduce the FLOPs (floating-point operations) in the feature channel fusion process. Mosaic data augmentation is also added to improve the feature expression performance. A public road disease dataset is reconstructed to verify the performance of the proposed method. The proposed model is trained and deployed to NVIDIA Jetson Nano for the experiment, and the results show that the average accuracy of the proposed model reaches 88.17%, increased by 4.01%, and the model FPS (frames per second) reaches 12.51, increased by 184% compared with the existing YOLOv5s. Case studies show that the proposed method satisfies the practical application requirements of pavement disease detection.

27 March 2023 in Multimedia Tools and Applications
Verified
Journal Article
Closed access
PDF file
JYJiana Yao
Jiana Yao
YCYinze Chu
Yinze Chu
XXXinjian Xiang
Xinjian Xiang
BHBingqiang Huang
Bingqiang Huang
WXWu Xiaoli
Wu Xiaoli

26 August 2021 in IET Cyber-Systems and Robotics
Verified
Journal Article
Open access
PDF file
BHBingqiang Huang
Bingqiang Huang
HLHaiping Lin
Haiping Lin
ZHZejun Hu
Zejun Hu
XXXinjian Xiang
Xinjian Xiang
JYJiana Yao
Jiana Yao
YOLO (You Only Look Once), as a target detection algorithm with good speed and precision, is widely used in the industry. In the process of driving, the vehicle image captured by the driving camera is detected and it extracts the license plate and the front part of the vehicle. Compared with the network structure of YOLOv3‐tiny algorithm, the acquisition method of anchor box is improved by combining the Birch algorithm. In order to improve the real‐time performance, the original two‐scale detection is added to the multi‐scale prediction of three‐scale detection to ensure its accuracy. Finally, the experimental results show that the improved YOLOv3‐tiny network structure proposed in this study can improve the performance of mean‐average‐precision, intersection over union and speed by 5.99%, 17.52% and 48.4%, respectively, and the algorithm has certain robustness.

28 July 2023 in Applied Sciences
Verified
Journal Article
Open access
PDF file
BHBingqiang Huang
Bingqiang Huang
善吴善宝 吴
善宝 吴
XXXinjian Xiang
Xinjian Xiang
ZFZhengshun Fei
Zhengshun Fei
STShaohua Tian
Shaohua Tian
YWYunlong Weng
Yunlong Weng
This paper proposes an improved model based on YOLOv5s, specifically designed to overcome the challenges faced by current target detection algorithms in the field of electric bike helmet detection. In order to enhance the model’s ability to detect small targets and densely populated scenes, a specialized layer dedicated to small target detection and a novel loss function called Normalized Wasserstein Distance (NWD) are introduced. In order to solve the problem of increasing model parameters and complexity due to the inclusion of a small target detection layer, a Cross-Stage Partial Channel Mixing (CSPCM) on top of Convmix is designed. The collaborative fusion of CSPCM and the Deep Feature Consistency (DFC) attention mechanism makes it more suitable for hardware devices. In addition, the conventional Nearest Upsample technology is replaced with the advanced CARAFE Upsample module, further improving the accuracy of the model. Through rigorous experiments on carefully constructed datasets, the results show significant improvements in various evaluation indicators such as precision, recall, mAP.5, and mAP.95. Compared with the unmodified YOLOv5s algorithm, the proposed enhanced model achieves significant improvements of 1.1%, 8.4%, 5.2%, and 8.6% on these indicators, respectively, and these enhancements are accompanied by a reduction of 778,924 parameters. The experimental results on our constructed dataset demonstrate the superiority of the improved model and elucidate its potential applications. Furthermore, promising improvements for future research are suggested. This study introduces an efficient approach for improving the detection of electric bike helmets and verifies the effectiveness and practicality of the model through experiments. Importantly, the proposed scheme has implications for other target detection algorithms, especially in the field of small target detection.