Jihua Zhu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Jihua Zhu has a total of 29 co-authors. Here are the top co-authors by number of shared publications:
Liang Ma
Institute of Artificial Intelligence and Robotics, Xi'an Jiaotong University, P. R. China
3 shared publicationsShaoyi Du
Institute of Artificial Intelligence and Robotics, Xi'an Jiaotong University, P. R. China
2 shared publications- Ce Li
School of Automation and Electrical Engineering, Lanzhou University of Technology, Lanzhou 730050, China
1 shared publication - Chongyu Wang
Laboratory for Artificial Intelligence ( Xi'an Jiaotong University) 710049 Xi'an China
1 shared publication - Danchen Zhao
Institute of Artificial Intelligence and Robotics, Xi’an Jiaotong University, No. 28 Xianning West Road, Xi’an 710049, Shaanxi, China
1 shared publication Deyu Meng
Institute for Information and System Sciences, Xi'an Jiaotong University Xi'an 710049 People's Republic of China
1 shared publication
Latest Publications
- Open Access
OpenUrban3D: Annotation-Free Open-Vocabulary Semantic Segmentation of Large-Scale Urban Point Clouds
Open-vocabulary semantic segmentation enables models to recognize and segment objects from arbitrary natural language descriptions, offering the flexibility to handle novel, fine-grained, or functionally defined categories beyond fixed label sets. While this capability is crucial for large-scale urban point clouds that support applications such as digital twins, smart city management, and urban analytics, it remains largely unexplored in this domain. The main obstacles are the frequent absence of high-quality, well-aligned multi-view imagery in large-scale urban point cloud datasets and the poor generalization of existing three-dimensional (3D) segmentation pipelines across diverse urban environments with substantial variation in geometry, scale, and appearance. To address these challenges, we present OpenUrban3D, the first 3D open-vocabulary semantic segmentation framework for large-scale urban scenes that operates without aligned multi-view images, pre-trained point cloud segmentation networks, or manual annotations. Our approach generates robust semantic features directly from raw point clouds through multi-view, multi-granularity rendering, mask-level vision-language feature extraction, and sample-balanced fusion, followed by distillation into a 3D backbone model. This design enables zero-shot segmentation for arbitrary text queries while capturing both semantic richness and geometric priors. Extensive experiments on large-scale urban benchmarks, including SensatUrban and SUM, show that OpenUrban3D achieves significant improvements in both segmentation accuracy and cross-scene generalization over existing methods, demonstrating its potential as a flexible and scalable solution for 3D urban scene understanding.
Open AccessOpenUrban3D: Annotation-Free Open-Vocabulary Semantic Segmentation of Large-Scale Urban Point Clouds
Open-vocabulary semantic segmentation enables models to recognize and segment objects from arbitrary natural language descriptions, offering the flexibility to handle novel, fine-grained, or functionally defined categories beyond fixed label sets. While this capability is crucial for large-scale urban point clouds that support applications such as digital twins, smart city management, and urban analytics, it remains largely unexplored in this domain. The main obstacles are the frequent absence of high-quality, well-aligned multi-view imagery in large-scale urban point cloud datasets and the poor generalization of existing three-dimensional (3D) segmentation pipelines across diverse urban environments with substantial variation in geometry, scale, and appearance. To address these challenges, we present OpenUrban3D, the first 3D open-vocabulary semantic segmentation framework for large-scale urban scenes that operates without aligned multi-view images, pre-trained point cloud segmentation networks, or manual annotations. Our approach generates robust semantic features directly from raw point clouds through multi-view, multi-granularity rendering, mask-level vision-language feature extraction, and sample-balanced fusion, followed by distillation into a 3D backbone model. This design enables zero-shot segmentation for arbitrary text queries while capturing both semantic richness and geometric priors. Extensive experiments on large-scale urban benchmarks, including SensatUrban and SUM, show that OpenUrban3D achieves significant improvements in both segmentation accuracy and cross-scene generalization over existing methods, demonstrating its potential as a flexible and scalable solution for 3D urban scene understanding.
- 27 January 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Knowledge and Data Engineering
- Vol. PP (10414347) , 1
Joint Reason Generation and Rating Prediction for Explainable Recommendation
Most recommendation systems focus on predicting rating or finding aspect information in reviews to understand user preferences and item properties. However, these methods ignore the effectiveness and persuasiveness of recommendation results. Consequently, explainable recommendation, namely providing recommendation results with recommendation reasons at the same time, has attracted increasing attention of researchers due to its ability in fostering transparency and trust. It is lucky that some E-commerce websites provide a kind of new interaction box called Tips and users can express their comments on items with a simple sentence. This brings us an opportunity to realize explainable recommendation. Under the supervision of two explicit feedback, namely rating and textual tips, we can implement a multi-task learning model which can provide recommendation results and generate recommendation reasons at the same time. In this paper, we propose an Encoder-Decoder and Multi-Layer Perception (MLP) based Explainable Recommendation model named EMER to simultaneously implement reason generation and rating prediction. Items title contains significant product-related information and plays an important role in grabbing users attention, so we fuse it in our model to generate recommendation reasons. Numerous experiments on benchmark datasets demonstrate that our model is superior to the state-of-the-art models.
- 27 January 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Knowledge and Data Engineering
- Vol. PP (10414347) , 1
Joint Reason Generation and Rating Prediction for Explainable Recommendation
Most recommendation systems focus on predicting rating or finding aspect information in reviews to understand user preferences and item properties. However, these methods ignore the effectiveness and persuasiveness of recommendation results. Consequently, explainable recommendation, namely providing recommendation results with recommendation reasons at the same time, has attracted increasing attention of researchers due to its ability in fostering transparency and trust. It is lucky that some E-commerce websites provide a kind of new interaction box called Tips and users can express their comments on items with a simple sentence. This brings us an opportunity to realize explainable recommendation. Under the supervision of two explicit feedback, namely rating and textual tips, we can implement a multi-task learning model which can provide recommendation results and generate recommendation reasons at the same time. In this paper, we propose an Encoder-Decoder and Multi-Layer Perception (MLP) based Explainable Recommendation model named EMER to simultaneously implement reason generation and rating prediction. Items title contains significant product-related information and plays an important role in grabbing users attention, so we fuse it in our model to generate recommendation reasons. Numerous experiments on benchmark datasets demonstrate that our model is superior to the state-of-the-art models.
- 7 July 2020
- Published by Elsevier in Neurocomputing
- Vol. 413, 230-239
Adaptive weighted motion averaging with low-rank sparse for robust multi-view registration
- 7 July 2020
- Published by Elsevier in Neurocomputing
- Vol. 413, 230-239
Adaptive weighted motion averaging with low-rank sparse for robust multi-view registration
- Open Access
Road Scene Simulation Based on Vehicle Sensors: An Intelligent Framework Using Random Walk Detection and Scene Stage Reconstruction
Road scene model construction is an important aspect of intelligent transportation system research. This paper proposes an intelligent framework that can automatically construct road scene models from image sequences. The road and foreground regions are detected at superpixel level via a new kind of random walk algorithm. The seeds for different regions are initialized by trapezoids that are propagated from adjacent frames using optical flow information. The superpixel level region detection is implemented by the random walk algorithm, which is then refined by a fast two-cycle level set method. After this, scene stages can be specified according to a graph model of traffic elements. These then form the basis of 3D road scene models. Each technical component of the framework was evaluated and the results confirmed the effectiveness of the proposed approach.
Open AccessRoad Scene Simulation Based on Vehicle Sensors: An Intelligent Framework Using Random Walk Detection and Scene Stage Reconstruction
Road scene model construction is an important aspect of intelligent transportation system research. This paper proposes an intelligent framework that can automatically construct road scene models from image sequences. The road and foreground regions are detected at superpixel level via a new kind of random walk algorithm. The seeds for different regions are initialized by trapezoids that are propagated from adjacent frames using optical flow information. The superpixel level region detection is implemented by the random walk algorithm, which is then refined by a fast two-cycle level set method. After this, scene stages can be specified according to a graph model of traffic elements. These then form the basis of 3D road scene models. Each technical component of the framework was evaluated and the results confirmed the effectiveness of the proposed approach.
- 8 June 2017
- Published by Cambridge University Press (CUP) in Bulletin of the Australian Mathematical Society
- Vol. 96 (3) , 353-360
TWO NEW LOWER BOUNDS FOR THE SPARK OF A MATRIX
The $l_{0}$ -minimisation problem has attracted much attention in recent years with the development of compressive sensing. The spark of a matrix is an important measure that can determine whether a given sparse vector is the solution of an $l_{0}$ -minimisation problem. However, its calculation involves a combinatorial search over all possible subsets of columns of the matrix, which is an NP-hard problem. We use Weyl’s theorem to give two new lower bounds for the spark of a matrix. One is based on the mutual coherence and the other on the coherence function. Numerical examples are given to show that the new bounds can be significantly better than existing ones.- 8 June 2017
- Published by Cambridge University Press (CUP) in Bulletin of the Australian Mathematical Society
- Vol. 96 (3) , 353-360
TWO NEW LOWER BOUNDS FOR THE SPARK OF A MATRIX
The $l_{0}$ -minimisation problem has attracted much attention in recent years with the development of compressive sensing. The spark of a matrix is an important measure that can determine whether a given sparse vector is the solution of an $l_{0}$ -minimisation problem. However, its calculation involves a combinatorial search over all possible subsets of columns of the matrix, which is an NP-hard problem. We use Weyl’s theorem to give two new lower bounds for the spark of a matrix. One is based on the mutual coherence and the other on the coherence function. Numerical examples are given to show that the new bounds can be significantly better than existing ones.- 20 March 2015
- Published by Cambridge University Press (CUP) in Robotica
- Vol. 34 (11) , 2516-2531
Merging grid maps of different resolutions by scaling registration
SUMMARY: This paper considers the problem of merging grid maps that have different resolutions. Because the goal of map merging is to find the optimal transformation between two partially overlapping grid maps, it can be viewed as a special image registration issue. To address this special issue, the solution considers the non-common areas and designs an objective function based on the trimmed mean-square error (MSE). The trimmed and scaling iterative closest point (TsICP) algorithm is then proposed to solve this well-designed objective function. As the TsICP algorithm can be proven to be locally convergent in theory, a good initial transformation should be provided. Accordingly, scale-invariant feature transform (SIFT) features are extracted for the maps to be potentially merged, and the random sample consensus (RANSAC) algorithm is employed to find the geometrically consistent feature matches that are used to estimate the initial transformation for the TsICP algorithm. In addition, this paper presents the rules for the fusion of the grid maps based on the estimated transformation. Experimental results carried out with publicly available datasets illustrate the superior performance of this approach at merging grid maps with respect to robustness and accuracy.
- 20 March 2015
- Published by Cambridge University Press (CUP) in Robotica
- Vol. 34 (11) , 2516-2531
Merging grid maps of different resolutions by scaling registration
SUMMARY: This paper considers the problem of merging grid maps that have different resolutions. Because the goal of map merging is to find the optimal transformation between two partially overlapping grid maps, it can be viewed as a special image registration issue. To address this special issue, the solution considers the non-common areas and designs an objective function based on the trimmed mean-square error (MSE). The trimmed and scaling iterative closest point (TsICP) algorithm is then proposed to solve this well-designed objective function. As the TsICP algorithm can be proven to be locally convergent in theory, a good initial transformation should be provided. Accordingly, scale-invariant feature transform (SIFT) features are extracted for the maps to be potentially merged, and the random sample consensus (RANSAC) algorithm is employed to find the geometrically consistent feature matches that are used to estimate the initial transformation for the TsICP algorithm. In addition, this paper presents the rules for the fusion of the grid maps based on the estimated transformation. Experimental results carried out with publicly available datasets illustrate the superior performance of this approach at merging grid maps with respect to robustness and accuracy.
- Open Access
- 1 October 2014
- Published by Institution of Engineering and Technology (IET) in IET Image Processing
- Vol. 8 (10) , 582-590
Robust registration of partially overlapping point sets via genetic algorithm with growth operator
Recently, genetic algorithm (GA) has been introduced as an effective method to solve the registration problem. It maintains a population of candidate solutions for the problem and evolves by iteratively applying a set of stochastic operators. Accordingly, a key question is how to reduce the population size. In this study, the authors present two techniques for reducing the population size in the GA for registration of partially overlapping point sets. Based on the trimmed iterative closest point algorithm, they introduce a growth operator into the GA. The growth operator, which is also inspired by the biological evolution, can improve the GA efficiency for registration. Furthermore, they present a technique called centre alignment to confirm the value range of all the registration parameters, which can reduce the search space and allow the well‐designed GA to directly solve the registration problem. Experimental results carried out with the m ‐dimensional point sets illustrate its advantages over previous approaches.
- 1 October 2014
- Published by Institution of Engineering and Technology (IET) in IET Image Processing
- Vol. 8 (10) , 582-590
Open AccessRobust registration of partially overlapping point sets via genetic algorithm with growth operator
Recently, genetic algorithm (GA) has been introduced as an effective method to solve the registration problem. It maintains a population of candidate solutions for the problem and evolves by iteratively applying a set of stochastic operators. Accordingly, a key question is how to reduce the population size. In this study, the authors present two techniques for reducing the population size in the GA for registration of partially overlapping point sets. Based on the trimmed iterative closest point algorithm, they introduce a growth operator into the GA. The growth operator, which is also inspired by the biological evolution, can improve the GA efficiency for registration. Furthermore, they present a technique called centre alignment to confirm the value range of all the registration parameters, which can reduce the search space and allow the well‐designed GA to directly solve the registration problem. Experimental results carried out with the m ‐dimensional point sets illustrate its advantages over previous approaches.
- 1 July 2014
- Published by Science China Press., Co. Ltd. in SCIENTIA SINICA Informationis
- Vol. 44 (7) , 886-899
基于粒子滤波的部分对应点集刚体配准算法
To register point sets with outliers including noises and missing data, this paper proposes a novel rigid registration algorithm for partially overlapping point sets with particle filter. It first transforms the special registration problem into a posterior estimation problem, then adopts the motion model based on the alignment error and designs the effective observation model with iterative closest point algorithm. Finally, it directly uses particle filter to solve the rigid partially registration problem. Experimental results illustrate the superiority of the proposed algorithm over previous approaches.
- 1 July 2014
- Published by Science China Press., Co. Ltd. in SCIENTIA SINICA Informationis
- Vol. 44 (7) , 886-899
基于粒子滤波的部分对应点集刚体配准算法
To register point sets with outliers including noises and missing data, this paper proposes a novel rigid registration algorithm for partially overlapping point sets with particle filter. It first transforms the special registration problem into a posterior estimation problem, then adopts the motion model based on the alignment error and designs the effective observation model with iterative closest point algorithm. Finally, it directly uses particle filter to solve the rigid partially registration problem. Experimental results illustrate the superiority of the proposed algorithm over previous approaches.
- Open Access
- 1 September 2013
- Published by Institution of Engineering and Technology (IET) in Electronics Letters
- Vol. 49 (20) , 1267-1268
Efficient scaling registration algorithm for partially overlapping point sets
An efficient iterative closest point (ICP) variant is proposed for solving the scaling registration of partially overlapping point sets. By adopting the backward distance measurement and the overlapping rate, the aforementioned registration problem can be formulated as a least squares function, which can be solved in the iterative manner, inspired by the ICP algorithm. During each iteration, the inverse correspondences are set up to compute the overlapping percentage, scale and rigid transformation simultaneously between two m ‐D point sets. The experimental results clearly demonstrate its superior performances over previous approaches.
- 1 September 2013
- Published by Institution of Engineering and Technology (IET) in Electronics Letters
- Vol. 49 (20) , 1267-1268
Open AccessEfficient scaling registration algorithm for partially overlapping point sets
An efficient iterative closest point (ICP) variant is proposed for solving the scaling registration of partially overlapping point sets. By adopting the backward distance measurement and the overlapping rate, the aforementioned registration problem can be formulated as a least squares function, which can be solved in the iterative manner, inspired by the ICP algorithm. During each iteration, the inverse correspondences are set up to compute the overlapping percentage, scale and rigid transformation simultaneously between two m ‐D point sets. The experimental results clearly demonstrate its superior performances over previous approaches.