Pei Yu
Pei Yu has a total of 41 co-authors. Here are the top co-authors by number of shared publications:
- Jing Li
Beijing Institute of Petrochemical Technology, Beijing, China
5 shared publications - Lili Zhang
Beijing Institute of Petrochemical Technology, Beijing, China
5 shared publications - Wei Wei
Beijing Institute of Petrochemical Technology, Beijing, China
4 shared publications - Fang Wang
College of Information Engineering, Beijing Institute of Petrochemical Technology, Beijing 102617, China
3 shared publications - Kang Yang
Rocket Force University of Engineering, Xi’an 710025, China
3 shared publications Lili Zhang
Beijing Institute of Petrochemical Technology, Beijing, China
3 shared publications
Latest Publications
- Open Access
- 19 January 2026
- Published by Springer Nature in Journal of King Saud University Computer and Information Sciences
- Vol. 38 (2)
HBD-NanoDetPlus:a lightweight detection model for vehicle high beam headlights in rainy and snowy weather
- 19 January 2026
- Published by Springer Nature in Journal of King Saud University Computer and Information Sciences
- Vol. 38 (2)
Open AccessHBD-NanoDetPlus:a lightweight detection model for vehicle high beam headlights in rainy and snowy weather
- 20 August 2025
- Published by Oxford University Press (OUP) in The Computer Journal
- Vol. 69 (1) , 63-73
Clairaudience: a lightweight attentional residual neural network with data augmentation and feature fusion for underwater acoustic target recognition
Marine engineering has boomed and many deep learning-based methods have been proposed for underwater acoustic target recognition. However, most of these methods are dedicated to develop more complex convolutional neural networks to achieve better performance. This results in these models being unable to be deployed to low cost and miniaturized automatic underwater vehicles. A novel lightweight attentional residual neural network with data augmentation and feature fusion is proposed in this paper. Mel Frequency Cepstral Coefficient (MFCC), delta-MFCC and delta–delta MFCC features are extracted in the time dimension for fusion to obtain the fusion feature. The SpecAugment data augmentation is also used to enhance the randomness and diversity of features by masking in time and frequency dimensions randomly. Shuffle attention in the residual blocks is introduced to enhance the representation of features. The lightweight model is evaluated and compared by using several metrics on ShipsEar and DeepShip datasets. The proposed lightweight model only requires 1.628 M parameters for the trained model. This work shows that the proposed method requires small memory storage, while it achieved comparative performance.
- 20 August 2025
- Published by Oxford University Press (OUP) in The Computer Journal
- Vol. 69 (1) , 63-73
Clairaudience: a lightweight attentional residual neural network with data augmentation and feature fusion for underwater acoustic target recognition
Marine engineering has boomed and many deep learning-based methods have been proposed for underwater acoustic target recognition. However, most of these methods are dedicated to develop more complex convolutional neural networks to achieve better performance. This results in these models being unable to be deployed to low cost and miniaturized automatic underwater vehicles. A novel lightweight attentional residual neural network with data augmentation and feature fusion is proposed in this paper. Mel Frequency Cepstral Coefficient (MFCC), delta-MFCC and delta–delta MFCC features are extracted in the time dimension for fusion to obtain the fusion feature. The SpecAugment data augmentation is also used to enhance the randomness and diversity of features by masking in time and frequency dimensions randomly. Shuffle attention in the residual blocks is introduced to enhance the representation of features. The lightweight model is evaluated and compared by using several metrics on ShipsEar and DeepShip datasets. The proposed lightweight model only requires 1.628 M parameters for the trained model. This work shows that the proposed method requires small memory storage, while it achieved comparative performance.
- Open Access
- 2 April 2025
- Published by Springer Nature in Scientific Reports
- Vol. 15 (1) , 1-14
Underwater vessel sound recognition based on multi-layer feature and attention mechanism
Vessel recognition based on hydroacoustic signals is an important research area. The marine environment is complex and variable, which makes the transmission and reception process of the signals have some random cases. At the same time, there are various interference and noise sources in the water, such as waves, underwater equipment, marine organisms, etc., which bring difficulties to the identification and analysis of vessel targets. This paper proposed a model named Emphasized Dimension Attention and Future Fusion-Time Delay Neural Network (EDAFF-TDNN). The model adjusts the weights of the feature map dynamically by learning the correlation between dimensions through Squeeze and Excitation Block (SE-Block), which enables the model to capture the contextual information, thus the model performance is improved. The mechanism of feature fusion is also introduced to extract multi-layer features to improve the feature representation capability. The attention mechanism is added on top of TDNN. By considering the differences of each feature dimension, it enables the model to focus on the key information when learning feature representations. Which improves the model performance in complex scenarios. In addition, experiments of the model on the ShipsEar dataset show a recognition accuracy of 98.2%.
- 2 April 2025
- Published by Springer Nature in Scientific Reports
- Vol. 15 (1) , 1-14
Open AccessUnderwater vessel sound recognition based on multi-layer feature and attention mechanism
Vessel recognition based on hydroacoustic signals is an important research area. The marine environment is complex and variable, which makes the transmission and reception process of the signals have some random cases. At the same time, there are various interference and noise sources in the water, such as waves, underwater equipment, marine organisms, etc., which bring difficulties to the identification and analysis of vessel targets. This paper proposed a model named Emphasized Dimension Attention and Future Fusion-Time Delay Neural Network (EDAFF-TDNN). The model adjusts the weights of the feature map dynamically by learning the correlation between dimensions through Squeeze and Excitation Block (SE-Block), which enables the model to capture the contextual information, thus the model performance is improved. The mechanism of feature fusion is also introduced to extract multi-layer features to improve the feature representation capability. The attention mechanism is added on top of TDNN. By considering the differences of each feature dimension, it enables the model to focus on the key information when learning feature representations. Which improves the model performance in complex scenarios. In addition, experiments of the model on the ShipsEar dataset show a recognition accuracy of 98.2%.
- Top Cited Papers
- 25 October 2024
- Published by Elsevier in Engineering Applications of Artificial Intelligence
- Vol. 139
TSD-DETR: A lightweight real-time detection transformer of traffic sign detection for long-range perception of autonomous driving
- 25 October 2024
- Published by Elsevier in Engineering Applications of Artificial Intelligence
- Vol. 139
Top Cited PapersTSD-DETR: A lightweight real-time detection transformer of traffic sign detection for long-range perception of autonomous driving
- Open Access
- 18 January 2024
- Published by Wiley in Journal of Sensors
- Vol. 2024, 1-10
Fire-PPYOLOE: An Efficient Forest Fire Detector for Real-Time Wild Forest Fire Monitoring
Forest fire has the characteristics of sudden and destructive, which threatens safety of people’s life and property. Automatic detection and early warning of forest fire in the early stage is very important for protecting forest resources and reducing disaster losses. Unmanned forest fire monitoring is one popular way of forest fire automatic detection. However, the actual forest environment is complex and diverse, and the vision image is affected by various factors easily such as geographical location, seasons, cloudy weather, day and night, etc. In this paper, we propose a novel fire detection method called Fire-PPYOLOE. We design a new backbone and neck structure leveraging large kernel convolution to capture a large arrange area of reception field based on the existing fast and accurate object detection model PP-YOLOE. In addition, our model maintains the high-speed performance of the single-stage detection model and reduces model parameters by using CSPNet significantly. Extensive experiments are conducted to show the effectiveness of Fire-PPYOLOE from the views of detection accuracy and speed. The results show that our Fire-PPYOLOE is able to detect the smoke- and flame-like objects because it can learn features around the object to be detected. It can provide real-time forest fire prevention and early detection.
- 18 January 2024
- Published by Wiley in Journal of Sensors
- Vol. 2024, 1-10
Open AccessFire-PPYOLOE: An Efficient Forest Fire Detector for Real-Time Wild Forest Fire Monitoring
Forest fire has the characteristics of sudden and destructive, which threatens safety of people’s life and property. Automatic detection and early warning of forest fire in the early stage is very important for protecting forest resources and reducing disaster losses. Unmanned forest fire monitoring is one popular way of forest fire automatic detection. However, the actual forest environment is complex and diverse, and the vision image is affected by various factors easily such as geographical location, seasons, cloudy weather, day and night, etc. In this paper, we propose a novel fire detection method called Fire-PPYOLOE. We design a new backbone and neck structure leveraging large kernel convolution to capture a large arrange area of reception field based on the existing fast and accurate object detection model PP-YOLOE. In addition, our model maintains the high-speed performance of the single-stage detection model and reduces model parameters by using CSPNet significantly. Extensive experiments are conducted to show the effectiveness of Fire-PPYOLOE from the views of detection accuracy and speed. The results show that our Fire-PPYOLOE is able to detect the smoke- and flame-like objects because it can learn features around the object to be detected. It can provide real-time forest fire prevention and early detection.
- Open Access
- 25 September 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 104806-104816
An Improved Autonomous Inertial-Based Integrated Navigation Scheme Based on Vehicle Motion Recognition
The autonomous navigation technology of INS has always attracted the attention of many scholars. In satellite-denied environment, making full use of the auxiliary sensors with autonomous characteristics can suppress the dynamic error generated by INS in autonomous navigation effectively. However, auxiliary sensors like an odometer (OD) and laser doppler velocimeter (LDV) will cause instability to the navigation system. To solve the above problems, we propose an improved autonomous inerital-based integrated navigation scheme utilizing vehicle motion recognition. The main research work can be summarized as: (1) Non-holonomic constrained velocity model and tri-axis zero velocity model has been established for inertial-based integrated navigation model. (2) Vehicle motion state recognition method has been proposed utilizing the IMU data. As shown in the experimental results, the proposed method can estimate the IMU biases and increase the positioning accuracy effectively as compared with the traditional methods.
- 25 September 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 104806-104816
Open AccessAn Improved Autonomous Inertial-Based Integrated Navigation Scheme Based on Vehicle Motion Recognition
The autonomous navigation technology of INS has always attracted the attention of many scholars. In satellite-denied environment, making full use of the auxiliary sensors with autonomous characteristics can suppress the dynamic error generated by INS in autonomous navigation effectively. However, auxiliary sensors like an odometer (OD) and laser doppler velocimeter (LDV) will cause instability to the navigation system. To solve the above problems, we propose an improved autonomous inerital-based integrated navigation scheme utilizing vehicle motion recognition. The main research work can be summarized as: (1) Non-holonomic constrained velocity model and tri-axis zero velocity model has been established for inertial-based integrated navigation model. (2) Vehicle motion state recognition method has been proposed utilizing the IMU data. As shown in the experimental results, the proposed method can estimate the IMU biases and increase the positioning accuracy effectively as compared with the traditional methods.
- Open Access
- 4 May 2022
- Published by Wiley in Journal of Sensors
- Vol. 2022, 1-18
Control Strategy of Frequent Overflow at Intersection Based on Remote Sensing of Unmanned Aerial Vehicle and Vehicle Trajectory Data
This paper proposed an effective method, the overflow control strategy, for identifying and controlling the frequent overflow of signalized intersections based on the unmanned aerial vehicle (UAV) remote sensing and vehicle trajectory data that provide multi-dimensional vehicle dynamic data. This strategy can effectively control the overflow intersection and restore normal operation as soon as possible. The strategy was implemented as follows. First, UAV remote sensing was used to quickly find the overflow precursor at the intersection. Second, by using moving trajectory data, the second-flow model of the road section was constructed, and overflow identification indexes were established. Third, considering the traffic demand of controlled intersection and non-overflow phase from the global perspective, the queue length model of the non-overflow phase based on moving trajectory data was established and overflow identification indexes were taken as the objective function to design overflow control strategy set. Finally, according to the sampling rate requirement of moving trajectory data and the field traffic survey, the Changping District in Beijing, which meets the requirements, is selected as the case study for simulation and verification. Usually, when overflow occurs, total delay time and the average number of stop cycles at intersections are relatively high. However, by taking the proposed method, overflow at intersections was identified quickly, and a control strategy was implemented immediately, which alleviated the overflow and eventually eliminated it. The results show that the method proposed in this paper can accurately detect the occurrence of overflow, restrain and eliminate the overflow eventually.
- 4 May 2022
- Published by Wiley in Journal of Sensors
- Vol. 2022, 1-18
Open AccessControl Strategy of Frequent Overflow at Intersection Based on Remote Sensing of Unmanned Aerial Vehicle and Vehicle Trajectory Data
This paper proposed an effective method, the overflow control strategy, for identifying and controlling the frequent overflow of signalized intersections based on the unmanned aerial vehicle (UAV) remote sensing and vehicle trajectory data that provide multi-dimensional vehicle dynamic data. This strategy can effectively control the overflow intersection and restore normal operation as soon as possible. The strategy was implemented as follows. First, UAV remote sensing was used to quickly find the overflow precursor at the intersection. Second, by using moving trajectory data, the second-flow model of the road section was constructed, and overflow identification indexes were established. Third, considering the traffic demand of controlled intersection and non-overflow phase from the global perspective, the queue length model of the non-overflow phase based on moving trajectory data was established and overflow identification indexes were taken as the objective function to design overflow control strategy set. Finally, according to the sampling rate requirement of moving trajectory data and the field traffic survey, the Changping District in Beijing, which meets the requirements, is selected as the case study for simulation and verification. Usually, when overflow occurs, total delay time and the average number of stop cycles at intersections are relatively high. However, by taking the proposed method, overflow at intersections was identified quickly, and a control strategy was implemented immediately, which alleviated the overflow and eventually eliminated it. The results show that the method proposed in this paper can accurately detect the occurrence of overflow, restrain and eliminate the overflow eventually.
- Open Access
- 16 March 2022
- Published by Springer Nature in Scientific Reports
- Vol. 12 (1) , 1-11
Research on integrated simulation platform for urban traffic control connecting simulation and practice
Though effective in theoretical simulation, the established traffic control models and optimization algorithms will result in model mismatch or even control strategy failure in actual application. However, they are commonly adopted in traffic signal control research, resulting in the unavailability of many exceptional control algorithms in practice. Simulation should function as a bridge between theoretical research and actual application, allowing the gap between the two to be communicated and made up for. However, an effective connection between the two has yet to be established to enable simulation methods in existing traffic control research. To this end, we designed and developed a simulation platform for "Online Application—HILS (Hardware-in-the-Loop Simulation)—Practice" integration over traffic signal control. In this paper, the architecture and characteristics of the integrated simulation platform were described. Besides, the function of each module of the platform was detailed, followed by listing simulation examples for six complex scenarios, with the active control scenario being selected for simulation comparison analysis. The findings demonstrated extensive road network simulation with the integrated simulation platform, multidimensional control variables, control strategies with support, as well as stable and reliable operation. It can be used to verify several sorts of traffic control simulation with variable dimensions.
- 16 March 2022
- Published by Springer Nature in Scientific Reports
- Vol. 12 (1) , 1-11
Open AccessResearch on integrated simulation platform for urban traffic control connecting simulation and practice
Though effective in theoretical simulation, the established traffic control models and optimization algorithms will result in model mismatch or even control strategy failure in actual application. However, they are commonly adopted in traffic signal control research, resulting in the unavailability of many exceptional control algorithms in practice. Simulation should function as a bridge between theoretical research and actual application, allowing the gap between the two to be communicated and made up for. However, an effective connection between the two has yet to be established to enable simulation methods in existing traffic control research. To this end, we designed and developed a simulation platform for "Online Application—HILS (Hardware-in-the-Loop Simulation)—Practice" integration over traffic signal control. In this paper, the architecture and characteristics of the integrated simulation platform were described. Besides, the function of each module of the platform was detailed, followed by listing simulation examples for six complex scenarios, with the active control scenario being selected for simulation comparison analysis. The findings demonstrated extensive road network simulation with the integrated simulation platform, multidimensional control variables, control strategies with support, as well as stable and reliable operation. It can be used to verify several sorts of traffic control simulation with variable dimensions.
- Open Access
- 1 October 2018
- Published by IOP Publishing in IOP Conference Series: Materials Science and Engineering
- Vol. 428 (1) , 012026
Design and Implementation of Inertial Measurement Information Acquisition System Based on XML
Platform inertial measurement information acquisition system is to acquire, store and display inertial measurement information of inertial platform navigation system by means of data communication. Due to the different types of platform inertial navigation system, the communication protocols are diverse. In this paper conducts the repetitive coding problem caused by different formats of communication protocols and put forward an improved method based on XML template applied to format of data unit in protocols. This method greatly improved the generality of acquisition software. In the premise that the acquisition board is not replaced, this method according to the type of platform to change the XML template, without compiled twice time. The inertial information acquisition system has been applied in practice. Practice has proved that this method can significantly reduce the workload of code maintenance and cut down the cost of personnel.
- 1 October 2018
- Published by IOP Publishing in IOP Conference Series: Materials Science and Engineering
- Vol. 428 (1) , 012026
Open AccessDesign and Implementation of Inertial Measurement Information Acquisition System Based on XML
Platform inertial measurement information acquisition system is to acquire, store and display inertial measurement information of inertial platform navigation system by means of data communication. Due to the different types of platform inertial navigation system, the communication protocols are diverse. In this paper conducts the repetitive coding problem caused by different formats of communication protocols and put forward an improved method based on XML template applied to format of data unit in protocols. This method greatly improved the generality of acquisition software. In the premise that the acquisition board is not replaced, this method according to the type of platform to change the XML template, without compiled twice time. The inertial information acquisition system has been applied in practice. Practice has proved that this method can significantly reduce the workload of code maintenance and cut down the cost of personnel.