Journal of Shandong University(Engineering Science)

Semantic segmentation method for potato common scab images based on multiscale feature fusion

WU Qiulan, SHANG Suya, ZHANG Jiahui, SUN Shouxin, ZHANG Feng, ZHOU Bo, GAO Zheng, SHI Wenchong

Journal of Shandong University(Engineering Science). 2025, 55(4): 1-8. doi:10.6040/j.issn.1672-3961.0.2024.232

Abstract ( 251 )

PDF (9072KB) ( 81 )

Save

References | Related Articles | Metrics

To precisely segment potato common scab lesions, a semantic segmentation model named MSFF-UNet was proposed. During the decoder's upward fusion process, feature enhancement was implemented through convolution and normalization operations to improve the differentiation of growth conditions in lesions of varying sizes. Additionally, multi-dimensional feature fusion capability was incorporated, where enhanced feature extraction from the decoder's high-level data was performed and subsequently fused with low-level data to capture semantic information of potatoes or common scab lesions at different scales. The results demonstrated that the improved semantic segmentation model achieved 93.90% precision, 93.51% mean class pixel accuracy, and 87.72% mean intersection over union, effectively enabling accurate segmentation of potatoes and common scab lesions.

Photovoltaic defect detection based on fine-grained feature enhancement and scale matching

SUO Daxiang, LI Bo

Journal of Shandong University(Engineering Science). 2025, 55(4): 9-17. doi:10.6040/j.issn.1672-3961.0.2024.170

Abstract ( 238 )

PDF (6558KB) ( 40 )

Save

References | Related Articles | Metrics

A high-precision defect detection method for photovoltaic panels was proposed to address the issue of missed and false detections associated with small defect targets in unmanned aerial vehicle(UAV)inspections of photovoltaic power stations. Considering the distribution characteristics of target sizes identified during UAV inspections, a strategy that combines fine-grained feature enhancement with scale matching was introduced to enhance the accuracy of small target defect detection in photovoltaics. Distinguished from traditional small target enhancement strategies including data augmentation, multi-scale learning, and feature enhancement, a detail-preserving semantic enhancement module was incorporated to retain fine-grained details and to mine related coarse-grained semantic details. A multi-scale detection strategy featuring anchor-prediction head matching was introduced to ensure the compatibility of anchor sizes with feature maps. The method achieved an average mean precision of 59.4% on the PVEL-AD dataset and 97.6% on the CARPK dataset, significantly improving the performance of photovoltaic defect target detection compared to mainstream object detection models.

Face image inpainting based on texture and structure interaction

ZHOU Zunfu, ZHANG Qian, SHI Jiliang, YUE Shiqin

Journal of Shandong University(Engineering Science). 2025, 55(4): 18-28. doi:10.6040/j.issn.1672-3961.0.2024.047

Abstract ( 232 )

PDF (12707KB) ( 38 )

Save

References | Related Articles | Metrics

Aiming at the issue of losing contextual semantic information when extracting deep features by learning-based face image inpainting methods, a generator with an efficient normalized attention mechanism was proposed, which extracted deep features from face images more effectively and better aggregated low-level and high-level features at multiple scales. To enhance the consistency of the generated images, a bi-level gated feature fusion module with residual main path transformation was introduced. This module further fused decoded texture and structure information, and incorporated an enhanced contextual feature aggregation module, in which an improved prompt generation block enabled prompt parameters to interact between features at multiple scales, guiding the dynamic adjustment of the inpainting network to generate realistic and believable face images. Experimental results on the CelebA-HQ datasets showed that this research method achieved 37.74 dB, 0.983 0, 0.24%, and 1.489 in terms of peak signal-to-noise ratio (R_PSN), structural similarity (S_SIM), mean absolute error(E_MA), and Fréchet inception distance (D_FI). On the LFW dataset, the R_PSN, S_SIM, E_MA, and D_FI of this research method achieved 39.19 dB, 0.987 7, 0.21%, and 3.555. Compared with five other mainstream methods, this research method achieved quite competitive results. Qualitative and quantitative experiments demonstrated that this research method could effectively restore corrupted facial structure and texture information.

Multi-granularity alignment network for image-text matching

WANG Xufeng, ZHOU Di, ZHANG Fenglei, SONG Xuemeng, LIU Meng

Journal of Shandong University(Engineering Science). 2025, 55(4): 29-39. doi:10.6040/j.issn.1672-3961.0.2024.024

Abstract ( 251 )

PDF (4315KB) ( 47 )

Save

References | Related Articles | Metrics

To precisely match image and text data, a multi-granularity alignment network(MGAN)was proposed. By adopting a contrastive language-image pre-training model and a Transformer-based bidirectional encoder model, MGAN extracted information at three different granularities: patch level, regional level, and global level, addressing the shortcomings of single-granularity information matching. A multi-level alignment mechanism was employed based on the characteristics of information at each level. At the regional level, a multi-view summarization module was integrated, allowing MGAN to effectively handle the one-to-many description problems between images and texts. At the patch level, a cross-modal similarity interaction modeling module was introduced to further enhance the detailed interactions between images and texts. Extensive experimental results on the publicly available datasets Flickr30K and MS-COCO demonstrated that MGAN achieved promising performance, confirming the effectiveness of the multi-granularity alignment network approach.

Industrial product surface defect detection based on self supervised convolution and parameter free attention mechanism

ZHOU Qunying, SUI Jiacheng, ZHANG Ji, WANG Hongyuan

Journal of Shandong University(Engineering Science). 2025, 55(4): 40-47. doi:10.6040/j.issn.1672-3961.0.2024.183

Abstract ( 262 )

PDF (3772KB) ( 34 )

Save

References | Related Articles | Metrics

Currently, surface defect detection in industrial products using deep learning faced several challenges. These included a limited number of negative sample datasets, susceptibility to complex industrial environments that hindered effective feature extraction, and high labor cost for labeling datasets. In order to solve the above problems,a knowledge distillation-based surface defect detection method for industrial products was proposed. The model incorporated a self-supervised predictive convolution module and a parameter-free attention mechanism. It transferred the rich feature knowledge learned by the teacher network to the student network. This approach effectively improved the model feature expression and localized defects at the pixel level. The model was experimented on the MVTec-AD dataset, and the comparison results with the state-of-the-art experimental methods showed that its detection and localization metrics were improved on A_ROC. The results proved that the method could improve the model's detection and localization capabilities.

Single image 3D model retrieval based on attention and view information

HAN Xiaofan, DIAO Zhenyu, ZHANG Chengyu, NIE Huijia, ZHAO Xiuyang, NIU Dongmei

Journal of Shandong University(Engineering Science). 2025, 55(4): 48-55. doi:10.6040/j.issn.1672-3961.0.2024.165

Abstract ( 206 )

PDF (2783KB) ( 24 )

Save

References | Related Articles | Metrics

To extract effective feature descriptors and reduce the significant differences between 2D images and 3D models, a method based on attention and view information was proposed. The method introduced a spatial attention mechanism into the model's feature extraction module to enhance the effectiveness of the model's feature descriptors. The 2D views of 3D models were incorporated into the process of learning query image features to reduce the domain gap between the image domain and the model domain. Experiments were conducted on three representative benchmark datasets: Pix3D, Comp Cars, and Stanford Cars. The results showed that the best retrieval accuracy improved by 5%. The proposed method effectively retrieved similar 3D models from a single image and improved the retrieval accuracy.

Review of knowledge distillation based on generative adversarial networks

YANG Jucheng, LU Kaikui, WANG Yuan

Journal of Shandong University(Engineering Science). 2025, 55(4): 56-71. doi:10.6040/j.issn.1672-3961.0.2024.055

Abstract ( 237 )

PDF (3155KB) ( 65 )

Save

References | Related Articles | Metrics

To summarize the application of generative adversarial networks in knowledge distillation, and explore the collaborative mechanisms and optimization potential of generative adversarial networks in knowledge distillation, a review of knowledge distillation based on generative adversarial networks was conducted. Research progress was reviewed in four categories of knowledge distillation, including methods based on output features, intermediate features, relational features, and structural features. The advantages and disadvantages of each approach were analyzed. The classification and development of knowledge distillation methods based on generative adversarial networks were introduced in detail. Limitations of these knowledge distillation techniques based on generative adversarial networks were identified, and potential directions for optimization and application expansion were proposed.

The multi-sensor fusion mapping and relocalization based on LVI-SAM-Stereo in indoor and outdoor scenes

JIANG Fengyang, CHENG Yao, HAN Zhe, WANG Huaizhen, ZHOU Fengyu, DONG Lei

Journal of Shandong University(Engineering Science). 2025, 55(4): 72-83. doi:10.6040/j.issn.1672-3961.0.2025.004

Abstract ( 365 )

PDF (16822KB) ( 68 )

Save

References | Related Articles | Metrics

Aiming at the problems of low mapping and relocalization accuracy, as well as poor scene adaptability, for robots in indoor and outdoor scenes, a tightly-coupled light detection and ranging(LiDAR)-visual-inertial odometry via smoothing, mapping, and relocalization by stereo(LVI-SAM-Stereo)method was proposed. The LiDAR-inertial pose estimation model was constructed by utilizing point-line and point-plane distances. Multi-sensor information interaction enabled rapid initialization of stereo-inertial odometry, with the odometry pose being optimized through reprojection error minimization. A cross-modal loop closure detection mechanism combining Scan-Context with visual features effectively reduced incorrect loop closures. A bidirectional relocalization architecture was developed, where factor graph-optimized odometry provided initial pose estimation for visual tracking, while perspective-n-point(PnP)-derived visual poses assisted LiDAR point cloud registration. A thorough evaluation with both datasets and real-world experiments verified that LVI-SAM-Stereo achieved 3.10% and 5.97% higher outdoor mapping accuracy compared to tightly-coupled LiDAR inertial odometry via smoothing and mapping(LIO-SAM)and tightly-coupled LiDAR-visual-inertial odometry via smoothing and mapping(LVI-SAM), respectively. Indoor average drift decreased by 72.7% and 43.05% versus these benchmarks. The system significantly improved mapping precision and scene adaptability. The relocalization satisfied the engineering requirements for autonomous navigation of robot products.

The spatial distribution of unsaturated soil and support design method of urban subway foundation pit

LI Lianxiang, GUO Longde, WANG Kunyi, WANG Peiyan, CHE Xiuxi, QIU Yefan

Journal of Shandong University(Engineering Science). 2025, 55(4): 84-92. doi:10.6040/j.issn.1672-3961.0.2024.015

Abstract ( 205 )

PDF (7782KB) ( 30 )

Save

References | Related Articles | Metrics

In order to promote the application of unsaturated soil theory in foundation pit support design, the thickness distribution of urban unsaturated soil was defined based on Jinan City. A simplified soil-water characteristic curve suitable for Jinan unsaturated soil was established by using matrix suction monitoring and unsaturated soil strength test data. Based on the total cohesion method, the formula of horizontal resistance coefficient and soil pressure of unsaturated soil in support structure expressed by soil saturation was derived, and the general design method of unsaturated soil support suitable for practical engineering application was obtained. The applicability of this method was demonstrated by using a subway foundation pit case. The case analysis showed that the internal force and displacement of the supporting structure calculated by the unsaturated soil theory were greatly reduced compared with the saturated soil theory, and were closer to the actual monitoring value. The research result showed that under the current engineering investigation level, the typical matrix suction measurement could be used to obtain and simplify soil-water characteristic curves in urban areas, establish the relationship between saturation and matrix suction, and complete the support design of unsaturated soil foundation pit, which could provide references for the design of urban unsaturated soil foundation pit.

The influence of mineral composition on uniaxial compression mechanical properties of granite based on PFC-GBM method

ZHANG Qinghao, MA Ruiyang, LIN Peng, XIE Huihui, WANG Zhaoyang, KANG Jintao, LOU Yanfei

Journal of Shandong University(Engineering Science). 2025, 55(4): 93-107. doi:10.6040/j.issn.1672-3961.0.2024.271

Abstract ( 225 )

PDF (26977KB) ( 34 )

Save

References | Related Articles | Metrics

In order to investigate the influence of three factors, namely the volume fraction, particle size and non-uniform distribution of various minerals in granite, on its compressive properties, the GBM(grain-based model)method was used to model the mineral composition of granite, to determine the strength of each factor's influence on compressive mechanical properties, and to reveal the mechanism of each factor's influence on compressive mechanical properties from the perspective of crack evolution. The results of the research showed that: each parameter had a high correlation with the strength of granite, and the granite compressive strength tended to increase with the growth of quartz and feldspar volume fraction and average grain size, and decrease with the increase of mica volume fraction and non-homogeneous factor; the influence of the mineral volume fraction parameter on the strength was in the order of strongest to weakest, which was the quartz, feldspar and mica, and the influence of the grain size parameter on the strength was in the following order average grain size, non-homogeneous factor; the correlation between the mineral volume fraction and the three types of intracrystalline contact ratio was high, the average grain size mainly affected the intracrystalline and inter-crystalline contact ratio, and the non-homogeneous factor led to change in the contact ratio of homogeneous and heterogeneous inter-crystalline contact, which were able to change the model rupture of the energy consumed and the path of crack development in varying degrees, and thus affected the uniaxial compressive strength.

Mechanical properties and constitutive relationship of steel slag fine aggregate concrete under impact load

XUE Gang, LIU Qiuyu, DONG Wei, LI Jingjun

Journal of Shandong University(Engineering Science). 2025, 55(4): 108-117. doi:10.6040/j.issn.1672-3961.0.2024.025

Abstract ( 214 )

PDF (10307KB) ( 14 )

Save

References | Related Articles | Metrics

In order to study the mechanical properties of steel slag fine aggregate concrete(SSC)under impact loading, axial impact compression tests of ordinary concrete and steel slag fine aggregate concrete with 10%, 20%, and 30% volume replacement ratios were carried out by using the 100 mm split Hopkinson pressure bar(SHPB). The effects of steel slag replacement ratio and strain rate on the dynamic compressive strength, impact toughness, dynamic increase factor(DIF)F_di, and failure mode of SSC were investigated. Based on the Z-W-T equation, a modified dynamic damage constitutive equation suitable for steel slag concrete was obtained. The results showed that SSC had a significant strain rate effect, with the same amount of steel slag, the dynamic compressive strength, F_di, impact toughness, and failure severity of SSC gradually increased with the growth of strain rate. The mechanical behavior under impact loading was changed by the addition of steel slag. When the strain rate was similar, the dynamic compressive strength and impact toughness of SSC showed an upward trend when the steel slag content increased from 0 to 30%, but the increase was relatively small, while the F_di value significantly decreased after the addition of steel slag. By fitting the experimental stress-strain curve, it could be concluded that the Z-W-T equation considering damage evolution could effectively describe the dynamic stress-strain relationship of SSC.

Experimental study on deformation characteristics of soil-structure interface under seepage conditions

LI Guang, LIU Jian, ZHOU Lizhi, LI Xiaohan, LÜ Gaohang, XIE Quanyi

Journal of Shandong University(Engineering Science). 2025, 55(4): 118-126. doi:10.6040/j.issn.1672-3961.0.2025.046

Abstract ( 190 )

PDF (8561KB) ( 19 )

Save

References | Related Articles | Metrics

In order to solve the problem that it was difficult to determine the deformation condition of the soil-structure interface under seepage, an experimental device for studying seepage-induced deformation at this interface was designed. This study investigated the distribution characteristics of interface deformation, soil surface deformation, and internal soil deformation, as well as their evolution with increasing hydraulic gradient under seepage action. The influence mechanism of different interface types and soil types on interface seepage failure was analyzed. The main conclusions were as follows:(1)the horizontal deformation of the contact surface was symmetrical in the test, the internal deformation of soil in the direction perpendicular to the contact surface showed a nonlinear decreasing trend with the increase of distance;(2)The seepage failure process of soil structure interface could be divided into three stages, stability, transition and failure, the deformation of the contact surface and the permeability coefficient of the contact surface showed the trend of linear increase, nonlinear increase and sharp increase in the three stages respectively;(3)The strains at the sand-structure interface were generally about 9.0%, strains at the silt-structure interface ranged from 7.9% to 9.1%, while strains at the low-liquid-limit clay-structure interface ranged from 5.2% to 6.5%.

Dynamic deformation law monitoring of extra-long span bridges based on GB-RAR technology

ZHANG Guojian, FU Lianlong, ZHANG Qingsong, SANG Wengang, LI Jianqiang, ZHOU Lu, FU Tao, LIU Shengzhen

Journal of Shandong University(Engineering Science). 2025, 55(4): 127-137. doi:10.6040/j.issn.1672-3961.0.2024.315

Abstract ( 220 )

PDF (12535KB) ( 23 )

Save

References | Related Articles | Metrics

As bridge spans increase, bridge flexibility is enhanced, making dynamic deflection deformation a critical indicator for assessing bridge health. Traditional monitoring methods, including sensors, measurement robots, and GPS, are limited in their ability to provide non-contact, high-frequency, and comprehensive dynamic deformation monitoring for large-span bridges. This research aimed to monitor the overall deformation patterns of the Yellow River Phoenix Bridge under dynamic vehicle loads using ground-based real aperture radar(GB-RAR)technology, with data quality enhanced through wavelet function analysis. The research results indicated that after wavelet denoising, the measurement error of GB-RAR was reduced to 0.016 5 mm, meeting the required accuracy for deformation monitoring. Under dynamic vehicle loads, bridge deformation followed the wave-sin-sqr model, with a maximum mid-span deflection deformation of 131.61 mm, which was within the tolerance specified by the General Codes for Highway Bridges and Culverts. These findings offered technological and data support for safety monitoring and reinforcement design of ultra-large-span bridges, such as the Yellow River Phoenix Bridge.

Research status and prospects of the industrial-agricultural collaborative development model based on a multi-dimensional benefit evaluation perspective

LIU Hao, JIANG Yuzhan, WANG Qingsong, LI Ziyang, DONG Yunlong, ZHANG Yujie,ZHANG Huibin

Journal of Shandong University(Engineering Science). 2025, 55(4): 138-148. doi:10.6040/j.issn.1672-3961.0.2024.344

Abstract ( 215 )

PDF (2505KB) ( 21 )

Save

References | Related Articles | Metrics

The current research on collaborative models based on multi-dimensional benefit evaluation lacked systematic assessment and frontier review, which restricted the comprehensive understanding of the model from the perspective of benefit evaluation. By defining the low-carbon development model of industry-agriculture collaboration, summarizing the achievements in current collaborative paths, research methods, effect evaluation, and other aspects based on literature research, cluster analysis, and other methods, the deficiencies in each research dimension were summarized and prospected. The study showed that the industry-agriculture collaboration model from the perspective of multi-dimensional benefits was characterized by "multiple paths, high potential and low risk", with deficiencies such as single path construction, weak integration in the research process, and weak dynamics in effect evaluation. Targeted development countermeasures and suggestions were proposed at three levels. At the path construction level, a multi-level three-dimensional collaborative path should be constructed. At the research method level, phased full-cycle research should be carried out. At the effect evaluation level, a dynamic evaluation system should be constructed. This study could provide useful academic references for the further deepening and expansion of research in this field and similar fields.

Model establishment and energy-saving analysis of liquid circulation heat recovery systems

LU Jiatong, ZHANG Chaoxu, DONG Xiaofei, ZHAO Hongxia, BAI Chao

Journal of Shandong University(Engineering Science). 2025, 55(4): 149-159. doi:10.6040/j.issn.1672-3961.0.2024.068

Abstract ( 208 )

PDF (4022KB) ( 32 )

Save

References | Related Articles | Metrics

To reduce the fresh air load, mathematical models of two liquid recirculating heat recovery systems were constructed based on flat finned circular tube heat exchangers. The model was divided into multiple treatment areas according to temperature and humidity based on outdoor meteorological parameters for 8 760 h in a year. The energy-saving effect of each system was calculated hour by hour. The results showed that under low humidity conditions, the performance of the three-coil and two-coil heat recovery systems was similar; under high humidity conditions, the three-coil system had a significant advantage, and its recovered energy was about 3.86 times of a two-coil system. Compared with the no heat recovery system, the annual energy-saving rate of the two-coil system was about 25.78%, while the three-coil system was as high as about 42.02%. The three-coil heat recovery system not only pre-cooled or pre-heated the outdoor air, but also met the reheating demand after dehumidification in summer, so the heat recovery effect was more stable throughout the year and the energy-saving performance was excellent. This study provided a theoretical basis for the modification of the heat recovery system of the air handling unit, which was of great significance for green building and sustainable development.

Study on associations between serum per- and polyfluoroalkyl substances levels and blood pressure in residents of Jinan

ZHANG Haoyu, XU Fei, LIU Yi, HOU Chengxi, DING Lei

Journal of Shandong University(Engineering Science). 2025, 55(4): 160-172. doi:10.6040/j.issn.1672-3961.0.2024.297

Abstract ( 181 )

Save

References | Related Articles | Metrics

Human was exposed to per- and polyfluoroalkyl substances(PFASs), which were implicated to be associated with elevated prevalence of hypertension. To evaluate the relationships between individual PFAS and PFAS mixture with blood pressure levels and hypertension risk, 18 PFASs in fasting serum samples collected from 326 individuals in Jinan, China were analyzed with an ultrahigh performance liquid chromatography system coupled with an Orbitrap mass spectrometer. Multivariable linear regression and logistic regression models were utilized to analyze the associations between individual PFAS and systolic blood pressure, diastolic blood pressure, and the risk of hypertension, respectively. To evaluate the joint effects of PFAS mixture, quantile g-computation and Bayesian kernel machine regression models were applied. All the models indicated a positive association between perfluorodecanoic acid mass concentration and diastolic blood pressure, a negative association between perfluorododecanoic acid mass concentration and diastolic blood pressure, and a positive association between perfluoroundecanoic acid mass concentration and risk of hypertension. According to a series of results from this study, it was concluded that both diastolic blood pressure and the risk of hypertension increased with the percentile of PFAS mixture mass concentration among the study population.

Table of Content