Yongzhuo Liu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Yongzhuo Liu has a total of 13 co-authors. Here are the top co-authors by number of shared publications:
Fayuan Xi
3 shared publicationsJiangmei Zhang
3 shared publicationsHaolin Liu
2 shared publications- Haolin Liu
School of Information Engineering, Southwest University of Science and Technology, Mianyang 621010, China
2 shared publications - Jiangmei Zhang
School of Information Engineering, Southwest University of Science and Technology, Mianyang 621010, China
2 shared publications Jing Lu
1 shared publication
Latest Publications
- Open Access
- 10 October 2026
- Published by Elsevier in SSRN Electronic Journal
Object-level response estimation for a collimated Geiger–Müller array under non-uniform contamination
Estimating individual object contributions from a collimated radiation array is difficult when extended objects have overlapping projections and non-uniform internal emission. We develop a geometry-informed inverse measurement method for a Geiger–Müller (GM) array and assess contribution estimates and their uncertainty under simulated readout conditions. OpenGATE response matrices retain the full material geometry. Four nonnegative subregion coefficients per object are estimated with a finite-window nonparalyzable count likelihood and within-object shrinkage, then aggregated into expected net responses before dead-time losses. A null-space condition characterizes identifiability of these aggregates, and scene-level split conformal calibration provides contribution intervals. Evaluation uses 1440 main-test and 2040 controlled observations from 40 independent test geometries. At 500 reference net counts, mean contribution-fraction error is 0.2653, 13.4% below uniform-source likelihood estimation overall and 35.9% lower under localized contamination. The uniform model remains more accurate for uniform contamination. Sensitivity-scaled intervals are 21.1% narrower on average than fixed-width intervals and jointly cover all object contributions in 40/40 test scenes. As a derived discrimination measure, the intervals identify a unique leading object in 55/360 observations. Source partitioning improves quantification when localized emission causes uniform-template mismatch; geometric response coupling and small contribution gaps still limit discrimination. The results characterize estimation performance for parameterized GM readout rather than a physically calibrated detector array.
- 10 October 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessObject-level response estimation for a collimated Geiger–Müller array under non-uniform contamination
Estimating individual object contributions from a collimated radiation array is difficult when extended objects have overlapping projections and non-uniform internal emission. We develop a geometry-informed inverse measurement method for a Geiger–Müller (GM) array and assess contribution estimates and their uncertainty under simulated readout conditions. OpenGATE response matrices retain the full material geometry. Four nonnegative subregion coefficients per object are estimated with a finite-window nonparalyzable count likelihood and within-object shrinkage, then aggregated into expected net responses before dead-time losses. A null-space condition characterizes identifiability of these aggregates, and scene-level split conformal calibration provides contribution intervals. Evaluation uses 1440 main-test and 2040 controlled observations from 40 independent test geometries. At 500 reference net counts, mean contribution-fraction error is 0.2653, 13.4% below uniform-source likelihood estimation overall and 35.9% lower under localized contamination. The uniform model remains more accurate for uniform contamination. Sensitivity-scaled intervals are 21.1% narrower on average than fixed-width intervals and jointly cover all object contributions in 40/40 test scenes. As a derived discrimination measure, the intervals identify a unique leading object in 55/360 observations. Source partitioning improves quantification when localized emission causes uniform-template mismatch; geometric response coupling and small contribution gaps still limit discrimination. The results characterize estimation performance for parameterized GM readout rather than a physically calibrated detector array.
- Open Access
Material-Aware First-Grasp Target Preselection for Mixed Rigid-Deformable Technical-Waste Mock-Ups
Robotic sorting of mixed rigid–deformable technical waste requires the selection of a graspable first target while minimizing disturbance to surrounding objects. This paper proposes a lightweight material-conditioned target-preselection method for cluttered RGB-D scenes. Object instances are detected using YOLO26n and segmented using SAM2-B, while their material categories are predicted using DPB-CNN. Depth-consistency filtering is subsequently applied to refine the candidate masks. Each candidate is evaluated according to geometric accessibility and local overlap, while visually inferred rigid-over-deformable relations are used to penalize or exclude deformable objects constrained by rigid-like objects. AnyGrasp generates a 6-DoF grasp pose using dense target points together with a downsampled workspace cloud retained for collision checking. Experiments were conducted on a UR5 platform using 40 predefined layouts covering four representative rigid–deformable interaction patterns, with three trials per method for each layout. The proposed method achieved a target selection accuracy (TSA) of 86.7%, a first-grasp success rate (FSR) of 80.8%, and a severe-disturbance rates (SDR) of 8.3%. After Holm correction, TSA was significantly higher than for both baselines, and SDR was significantly lower than for Native-AnyGrasp. The observed FSR improvements did not reach the corrected significance threshold. These results support improved first-grasp decision-making and reduced disturbance relative to Native-AnyGrasp in the evaluated technical-waste mock-up scenarios.
Open AccessMaterial-Aware First-Grasp Target Preselection for Mixed Rigid-Deformable Technical-Waste Mock-Ups
Robotic sorting of mixed rigid–deformable technical waste requires the selection of a graspable first target while minimizing disturbance to surrounding objects. This paper proposes a lightweight material-conditioned target-preselection method for cluttered RGB-D scenes. Object instances are detected using YOLO26n and segmented using SAM2-B, while their material categories are predicted using DPB-CNN. Depth-consistency filtering is subsequently applied to refine the candidate masks. Each candidate is evaluated according to geometric accessibility and local overlap, while visually inferred rigid-over-deformable relations are used to penalize or exclude deformable objects constrained by rigid-like objects. AnyGrasp generates a 6-DoF grasp pose using dense target points together with a downsampled workspace cloud retained for collision checking. Experiments were conducted on a UR5 platform using 40 predefined layouts covering four representative rigid–deformable interaction patterns, with three trials per method for each layout. The proposed method achieved a target selection accuracy (TSA) of 86.7%, a first-grasp success rate (FSR) of 80.8%, and a severe-disturbance rates (SDR) of 8.3%. After Holm correction, TSA was significantly higher than for both baselines, and SDR was significantly lower than for Native-AnyGrasp. The observed FSR improvements did not reach the corrected significance threshold. These results support improved first-grasp decision-making and reduced disturbance relative to Native-AnyGrasp in the evaluated technical-waste mock-up scenarios.
- Open Access
- 11 June 2026
- Published by Elsevier in SSRN Electronic Journal
OpenGATE-Based Gamma Response Mapping of a Lead-Collimated Multi-Range GM Detector Array for Nuclear Technical Waste Screening
Rapid screening of mixed nuclear technical waste requires a radiation measurement layer that can provide both conservative risk estimation and region-level spatial information. This study proposes an OpenGATE/Geant4-based gamma response mapping framework for a 10 × 10 planar multi-range Geiger–Mueller detector array with a lead-grid coarse collimation structure. The simulated photon-transport response is converted into a simulation-normalized dose-rate score map through a statistical multi-range GM fusion model considering dead-time correction, valid-range weighting, and counting uncertainty. The framework was evaluated using particle-number convergence, lead-grid parameter analysis, measurement-time assessment, source-depth and shielding robustness tests, known-level screening validation, independent randomized validation, and robotic sorting priority experiments. The lead grid reduced the neighbor-8/peak crosstalk ratio from 1.905 without lead to 1.075 for the 10 mm/40 mm configuration, with a 57.4% reduction in total counts. A 5 s acquisition achieved a repeated-postprocessing coefficient of variation of approximately 1.85%. The independent randomized validation achieved 96.0% accuracy with no dangerous underestimation. In 50 randomized waste-bag scenarios, the first selected item was an L3 or L4 source in 49/50 cases, with an average top-3 high-risk recall of 0.940. The reported values are simulation-normalized screening scores rather than calibrated ambient dose-equivalent rates; standard-source calibration and prototype validation are required before absolute dose-rate interpretation.
- 11 June 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessOpenGATE-Based Gamma Response Mapping of a Lead-Collimated Multi-Range GM Detector Array for Nuclear Technical Waste Screening
Rapid screening of mixed nuclear technical waste requires a radiation measurement layer that can provide both conservative risk estimation and region-level spatial information. This study proposes an OpenGATE/Geant4-based gamma response mapping framework for a 10 × 10 planar multi-range Geiger–Mueller detector array with a lead-grid coarse collimation structure. The simulated photon-transport response is converted into a simulation-normalized dose-rate score map through a statistical multi-range GM fusion model considering dead-time correction, valid-range weighting, and counting uncertainty. The framework was evaluated using particle-number convergence, lead-grid parameter analysis, measurement-time assessment, source-depth and shielding robustness tests, known-level screening validation, independent randomized validation, and robotic sorting priority experiments. The lead grid reduced the neighbor-8/peak crosstalk ratio from 1.905 without lead to 1.075 for the 10 mm/40 mm configuration, with a 57.4% reduction in total counts. A 5 s acquisition achieved a repeated-postprocessing coefficient of variation of approximately 1.85%. The independent randomized validation achieved 96.0% accuracy with no dangerous underestimation. In 50 randomized waste-bag scenarios, the first selected item was an L3 or L4 source in 49/50 cases, with an average top-3 high-risk recall of 0.940. The reported values are simulation-normalized screening scores rather than calibrated ambient dose-equivalent rates; standard-source calibration and prototype validation are required before absolute dose-rate interpretation.
- Open Access
- 5 October 2025
- Published by MDPI AG in Electronics
- Vol. 14 (19) , 3943
A Dual-Structured Convolutional Neural Network with an Attention Mechanism for Image Classification
This paper presents a dual-structured convolutional neural network (CNN) for image classification, which integrates two parallel branches: CNN-A with spatial attention and CNN-B with channel attention. The spatial attention module in CNN-A dynamically emphasizes discriminative regions by aggregating channel-wise information, while the channel attention mechanism in CNN-B adaptively recalibrates feature channel importance. The extracted features from both branches are fused through concatenation, enhancing the model’s representational capacity by capturing complementary spatial and channel-wise dependencies. Extensive experiments on a 12-class image dataset demonstrate the superiority of the proposed model over state-of-the-art methods, achieving 98.06% accuracy, 96.00% precision, and 98.01% F1-score. Despite a marginally longer training time, the model exhibits robust convergence and generalization, as evidenced by stable loss curves and high per-class recognition rates (>90%). The results validate the efficacy of dual attention mechanisms in improving feature discrimination for complex image classification tasks.
- 5 October 2025
- Published by MDPI AG in Electronics
- Vol. 14 (19) , 3943
Open AccessA Dual-Structured Convolutional Neural Network with an Attention Mechanism for Image Classification
This paper presents a dual-structured convolutional neural network (CNN) for image classification, which integrates two parallel branches: CNN-A with spatial attention and CNN-B with channel attention. The spatial attention module in CNN-A dynamically emphasizes discriminative regions by aggregating channel-wise information, while the channel attention mechanism in CNN-B adaptively recalibrates feature channel importance. The extracted features from both branches are fused through concatenation, enhancing the model’s representational capacity by capturing complementary spatial and channel-wise dependencies. Extensive experiments on a 12-class image dataset demonstrate the superiority of the proposed model over state-of-the-art methods, achieving 98.06% accuracy, 96.00% precision, and 98.01% F1-score. Despite a marginally longer training time, the model exhibits robust convergence and generalization, as evidenced by stable loss curves and high per-class recognition rates (>90%). The results validate the efficacy of dual attention mechanisms in improving feature discrimination for complex image classification tasks.