Liang Yu

Prof. Liang Yu

orcid

Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

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

Representative Publications

View All Publications
25 August 2022 in Mathematical Problems in Engineering
Verified
Journal Article
Open access
PDF file
FMFu Mo
Fu Mo
Liang YuLiang Yu
Liang Yu
ZZZhengkuan Zhang
Zhengkuan Zhang
YZYang Zhao
Yang Zhao
Aiming at the current problems in the safety monitoring of urban manhole covers, this paper proposes a safety monitoring system for manhole covers based on smart light poles. The system uses STM32F103C8T6 as the microcontroller, and processes and controls the movement, loss, tilt, flooding, and positioning data of the manhole cover. Then, the data frame is sent to the LoRa gateway of the nearby smart light pole through the LoRa communication protocol, and the LoRa gateway will transmit the abnormal state of the manhole cover to the large screen display of the smart light pole through the bus. At the same time, a voice broadcast is carried out to remind localized road vehicles and pedestrians. Besides, it is sent to the IoT cloud platform KitLink, and the cloud platform will push the abnormal status data to the subscribed management user, and the user will quickly process it according to the positioning data. Among them, the terminal sensing control device of the manhole cover adopts an integrated package, and at the same time, it is installed at the center point under the manhole cover by using a waterproof material. It can not only carry out reliable and accurate installation without changing the original components but also can effectively monitor the safety. Through the design and implementation test of the system, it can be seen that the system can efficiently and accurately realize all-round and multi-modal safety linkage monitoring of urban manhole covers.

1 January 2023 in Mathematical Biosciences and Engineering
Verified
Journal Article
Open access
PDF file
Liang YuLiang Yu
Liang Yu
ZZZhengkuan Zhang
Zhengkuan Zhang
YLYangbing Lai
Yangbing Lai
YZYang Zhao
Yang Zhao
FMFu Mo
Fu Mo
Unusual states of manhole covers (MCs), such as being tilted, lost or flooded, can present substantial safety hazards and risks to pedestrians and vehicles on the roadway. Most MCs are still being managed through manual regular inspections and have limited information technology integration. This leads to time-consuming and labor-intensive identification with a lower level of accuracy. In this paper, we propose an edge computing-based intelligent monitoring system for manhole covers (EC-MCIMS). Sensors detect the MC and send status and positioning information via LoRa to the edge gateway located on the nearby wisdom pole. The edge gateway utilizes a lightweight machine learning model, trained on the edge impulse (EI) platform, which can predict the state of the MC. If an abnormality is detected, the display and voice device on the wisdom pole will respectively show and broadcast messages to alert pedestrians and vehicles. Simultaneously, the information is uploaded to the cloud platform, enabling remote maintenance personnel to promptly repair and restore it. Tests were performed on the EI platform and in Dongguan townships, demonstrating that the average response time for identifying MCs is 4.81 s. Higher responsiveness and lower power consumption were obtained compared to cloud computing models. Moreover, the system utilizes a lightweight model that better reduces read-only memory (ROM) and random-access memory (RAM), while maintaining an average identification accuracy of 94%.

28 July 2023
Verified
Conference Paper
Closed access
FMFu Mo
Fu Mo
ZZZhengkuan Zhang
Zhengkuan Zhang
Liang YuLiang Yu
Liang Yu
Route planning is the main content of UAV flight, but the flight dynamics and geographical environment reduce the accuracy of route planning. The genetic algorithm to analyze route planning, improve the level of route planning the route planning time. Then, a comprehensive assessment of the drone flight is carried out. Finally, the GPS satellite system is used to verify the route planning and output the final route planning scheme. The results show that the improved genetic algorithm can accurately plan the route, and the accuracy is greater than 94%, which is improving the genetic algorithm can meet route planning requirements. The optimization degree of horizontal and vertical statistical features and states of the improved genetic algorithm is more than 90%, and the path planning time is 80 seconds, which is significantly better than the planning index of UAV's own system. Therefore, the improved genetic algorithm for UAV path planning is better and can improve the path effect.

12 April 2024
Verified
Conference Paper
Closed access
Liang YuLiang Yu
Liang Yu
YLYuting Lin
Yuting Lin
HXHaonan Xu
Haonan Xu
JYJiabao Yang
Jiabao Yang
GZGuotong Zhao
Guotong Zhao
Given that individuals spend more than 90% of their time indoors, research indicates a direct correlation between indoor air quality and overall health and well-being. With the rise of respiratory illnesses and increasing concerns about indoor air quality, traditional methods such as using potted plants, air purifiers, and basic air conditioning units with purification capabilities are insufficient for real-time monitoring. To address these challenges, this study proposes an intelligent indoor air quality monitoring system that leverages cloud-edge-end collaboration. The system includes a cost-effective air quality monitor on the device side to collect and localize indoor air quality data. This data is then transmitted to the edge-side ceiling light via WiFi-Mesh. A machine learning model deployed using the CNN-LSTM algorithm on the Edge Impluse platform is used to analyze and predict indoor air quality levels over time. The system uploads this information to a cloud MQTT server using the Cat.1 protocol and sends notifications to subscribed users via a mobile app, enabling individuals to monitor and adjust indoor air quality as needed. This proactive approach can help reduce air pollution, decrease disease incidence, and enhance overall quality of life. Physical and deployment tests demonstrate the system's high performance, with a training accuracy of 97.5%, F1-SCORE of 95%, and prediction accuracy of 94.1%, making it a reliable tool for indoor air quality management.

24 December 2025 in Electronics
Verified
Journal Article
Open access
PDF file
Liang YuLiang Yu
Liang Yu
XFXingcan Fang
Xingcan Fang
YZYuze Zeng
Yuze Zeng
WGWeili Guo
Weili Guo
达杨达 幸 杨
达 幸 杨
HSHengyi Sun
Hengyi Sun
The automation of quality control in agricultural food processing, particularly the detection of incomplete peeling in taro, constitutes a critical frontier for ensuring food safety and optimizing production efficiency in the Industry 4.0 era. However, this domain is fraught with significant technical challenges, primarily stemming from the inherent visual characteristics of residual peel: extremely minute scales relative to the vegetable body, highly irregular morphological variations, and the dense occlusion of objects on industrial conveyor belts. To address these persistent impediments, this study introduces a comprehensive solution comprising a specialized dataset and a novel detection architecture. We established the Taro Peel Industrial Dataset (TPID), a rigorously annotated collection of 18,341 high-density instances reflecting real-world production conditions. Building upon this foundation, we propose MDB-YOLO, a lightweight, multi-dimensional bionic detection model evolved from the YOLOv8s architecture. The MDB-YOLO framework integrates a synergistic set of innovations designed to resolve specific detection bottlenecks. To mitigate the conflict between background texture interference and tiny target detection, we integrated the C2f_EMA module with a Wise-IoU (WIoU) loss function, a combination that significantly enhances feature response to low-contrast residues while reducing the penalty on low-quality anchor boxes through a dynamic non-monotonic focusing mechanism. To effectively manage irregular peel shapes, a dynamic feature processing chain was constructed utilizing DySample for morphology-aware upsampling, BiFPN_Concat2 for weighted multi-scale fusion, and ODConv2d for geometric preservation. Furthermore, to address the issue of missed detections caused by dense occlusion in industrial stacking scenarios, Soft-NMS was implemented to replace traditional greedy suppression mechanisms. Experimental validation demonstrates the superiority of the proposed framework. MDB-YOLO achieves a mean Average Precision (mAP50-95) of 69.7% and a Recall of 88.0%, significantly outperforming the baseline YOLOv8s and advanced transformer-based models like RT-DETR-L. Crucially, the model maintains high operational efficiency, achieving an inference speed of 1.1 ms on an NVIDIA A100 and reaching 27 FPS on an NVIDIA Jetson Xavier NX using INT8 quantization. These findings confirm that MDB-YOLO provides a robust, high-precision, and cost-effective solution for real-time quality control in agricultural food processing, marking a significant advancement in the application of computer vision to complex biological targets.

4 November 2024
Verified
Conference Paper
Closed access
XZXiaochen Zhang
Xiaochen Zhang
KCKangji Cui
Kangji Cui
FMFu Mo
Fu Mo
Liang YuLiang Yu
Liang Yu
With the advancement of the new generation of technological revolution, the integration of AI technology into the field of education has deepened significantly, providing new opportunities for educational innovation. Inspired by the “5A” consumer experience path model in the digital marketing environment, this paper proposes a “5A” education model based on Artificial Intelligence Generated Content (AIGC) technology, aiming to establish a personalized and precise distance education environment through digital marketing concepts. The model is applied to the wireless sensor network course, facilitating comprehensive coverage from course information dissemination to the recording of students' learning behaviors across the five stages of Aware, Appeal, Ask, Action, and Advocate. This study employs AIGC technology to rapidly aggregate and generate multimodal educational content, thereby enhancing the visibility and attractiveness of the course while guiding students towards practical operations and knowledge sharing, ultimately improving both teaching effectiveness and the learning experience. Furthermore, this study evaluates and analyzes the AIGC+ “5A” education model by constructing online learning assessment and evaluation indicators. The results indicate that the model effectively enhances students' interest in learning, improves tailored teaching effectiveness and efficiency, and demonstrates its potential for application in the field of education.