Jianqing Wu’s research while affiliated with Shandong University and other places

... It presents a basic modeling problem based on LiDAR point clouds, which is only a small fragment of LiDAR systems applications; however, it is especially important in surveying, for example, in deformation analysis, inventory, and monitoring of engineering structures [15,31,32]. The methods addressed in the paper can probably be implemented into other LiDAR data processing, including classification, segmentation, tracking, etc., which have found many applications in robotics, autonomous vehicles, or generally computer-vision-based deployments [26,33,34]. ...

Reference:

Msplit Estimation with Local or Global Robustness Against Outliers—Applications and Limitations in LiDAR Data Processing

... Results demonstrated that the model could only capture certain advanced merging behavior. Ma et al. (2024) collected a merging trajectory database from roadside LiDAR sensors. A merging trajectory prediction model was established upon a self-supervised mechanism. ...

Reference:

Freeway Merging Trajectory Prediction for Automated Vehicles Using Naturalistic Driving Data

... This methodology avoids the limitations of traditional reinforcement learning techniques. It is robust to high-dimensional (large state-action space) tasks and issues related to a lack of generalization ability across diverse sets [20]. This novel strategy is nearly inestimable and probable for increasing the performance and dimensions of the scale of autonomous robots that may be employed in a long range. ...

Reference:

Clustering-Based Hierarchical Reinforcement Learning For Autonomous Robotics

... By constructing a self-learning adjacency matrix, the model facilitates the capture of hidden features of nodes and dynamic spatial correlations. Huo et al. [26] enhanced the road traffic flow prediction algorithm based on attention-based spatio-temporal graph convolutional network, it can effectively capture sudden changes in road traffic flow and diagnose instances of acute road traffic congestion. Sun et al. [27] proposed a Dual Dynamic Spatial-temporal graph convolution network (DDSTGCN), which not only models the dynamic property of the nodes of the traffic flow graph but also captures the dynamic spatial-temporal feature of the edges by transforming the traffic flow graph into its dual hypergraph. ...

Reference:

Spatio-temporal graph neural network based on time series periodic feature fusion for traffic flow prediction

... Biaxial compression tests on concrete after freeze-thaw cycles were conducted to analyze the dependence of ultimate strength on the number of freeze-thaw cycles and the relationship between the number of freeze-thaw cycles, the stress ratio, and the strain rate based on experimental data from Wang et al. [21]. Zhang et al. [22] investigated the compressive strength of foam concrete with different densities under the influence of single factors (such as water immersion, constant compression load, and freeze-thaw cycles) and coupled factors (such as water saturation constant load and constant load freeze-thaw cycles). They established predictive equations for compressive strength based on density, freeze-thaw cycles, and constant pressure, describing the strength variation of foam concrete under coupled loading. ...

Reference:

Mesoscopic Numerical Simulation of Freeze–Thaw Damage in Hydraulic Concrete

... Foam concrete, conventionally having a dry density of 300 to 1840 kg·m −3 [1], is a lightweight inorganic thermal insulating material with plenty of closed foams obtained by replacing lightweight aggregate with preprepared foams into the cement paste [2,3]. The elevated-density foam concrete can be used to mould load-bearing components like buffering pads or retaining walls, while low-density (less than 800 kg·m −3 ) foam concrete is usually employed for thermal insulation and noise absorption [4]. ...

Reference:

Mix proportion optimization and early strength development in modified foam concrete: an experimental study

... Methods with fixed reference objects primarily involve identifying known reference objects' shape, position, and other information to calculate the LiDAR's angular deviation, achieving the purpose of calibration. For example, Y. Park [13] proposed the use of a polygon plane identification board, Z. Pusztai [14] introduced a standard square marker board, and Zhu [15] employed a checkerboard marker board for calibration. However, these methods still have certain limitations for the angle calibration of roadside LiDAR point clouds. ...

Reference:

Research on an Adaptive Method for the Angle Calibration of Roadside LiDAR Point Clouds

... Most of the articles focused on detecting road damage or cracks. One article developed a Multi-level Attention Block (MLAB) mechanism using UAVs and YOLOv3 for pavement damage detection (Zhang et al. 2022). Another study developed a model named ARD-Unet that performed detection pixel-by-pixel. ...

Reference:

Applications of UAVs and AI in Civil Engineering: A Systematic Review of Current Uses and Future Directions in Road Construction

... Jaithavil D et al [20] used Transfer Learning to train three network models VGG16, InceptionV3 and MobileNetV2 on more than 1200 paddy seed datasets with overall recognition rates of 80.00%, 83.33% and 83.33% respectively. Zhou Y et al [21] added attention modules such as SELayer [22], EcaLayer [23], CBAM [24] and CoordAtt to YoloV5s and after experiments, it was found that YoloV5s with the addition of the CoordAtt attention had the best performance. Zhao W et al [25] introduced the CoordAtt attention mechanism to YOLOv5s as well as replacing the backbone module in the network with the convolutional blocks in the RepVGG [26] block structure to detect the flowering period of yellow chrysanthemums, and the final average accuracy reached 93.9%. ...

Reference:

IRB-5-CA Net:A Lightweight, Deep Learning-based Approach to Wheat Seed Identification

... YOLOv3 uses a more complex backbone network, referencing ResNet101 [21], but with faster execution speed. YOLOv5, compared to YOLOv4, introduces data augmentation techniques such as scaling, color space adjustment, and mosaic augmentation [22], and incorporates adaptive anchor boxes. The current trend in one-stage object detection is to build backbone networks with enhanced representation capabilities to improve algorithm accuracy. ...

Reference:

DFP-YOLO: a lightweight machine tool workpiece defect detection algorithm based on computer vision