HLHaijian Liu

Haijian Liu

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Haijian Liu has a total of 34 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    On the Robustness of Individual Tree Segmentation to Data Adversarial Attacks from Remote Sensing Point Clouds

    by Renhao Shen, Yongwei Miao, Haijian Liu

    Forests play a vital role in maintaining ecological balance, making accurate forest monitoring technologies essential. Remote sensing point cloud data always capture distinctive geometric features of forests, including the cylindrical symmetry of tree trunks and the radial symmetry of canopies. However, the inherent complexity of point cloud data, combined with their vulnerabilities to adversarial attacks, often disrupts these symmetrical patterns, significantly limiting the practical application of deep learning models in forest monitoring. This research presents a novel approach to enhance the robustness of individual tree segmentation networks by combining data augmentation and adversarial training techniques. Our method employs the FGSM algorithm and Gaussian noise attack to generate adversarial samples while utilizing data denoising and controlled noise injection for data augmentation. A dynamic adversarial training framework can adaptively adjust the proportion of adversarial samples during the network training stage to optimize the model. Using remote sensing point cloud datasets from Wisconsin, the experimental results demonstrate the effectiveness of the individual tree segmentation networks, PointNet++ and DBSCAN, in reducing attack success rates whilst improving the stability and accuracy of segmentation results under various adversarial conditions. This study highlights the potential for more robust forest monitoring systems capable of maintaining accuracy even when faced with data perturbations or intentional interference.

  • Open Access

    Comparison of UAV-Based LiDAR and Photogrammetric Point Cloud for Individual Tree Species Classification of Urban Areas

    by Qixia Man, Xinming Yang, Haijian Liu, Baolei Zhang, Pinliang Dong, Jingru Wu, Chunhui Liu, Changyin Han, Cong Zhou, Zhuang Tan, et al.

    UAV LiDAR and digital aerial photogrammetry (DAP) have shown great performance in forest inventory due to their advantage in three-dimensional information extraction. Many studies have compared their performance in individual tree segmentation and structural parameters extraction (e.g. tree height). However, few studies have compared their performance in tree species classification. Therefore, we have compared the performance of UAV LiDAR and DAP-based point clouds in individual tree species classification with the following steps: (1) Point cloud data processing: Denoising, smoothing, and normalization were conducted on LiDAR and DAP-based point cloud data separately. (2) Feature extraction: Spectral, structural, and texture features were extracted from the pre-processed LiDAR and DAP-based point cloud data. (3) Individual tree segmentation: The marked watershed algorithm was used to segment individual trees on canopy height models (CHM) derived from LiDAR and DAP data, respectively. (4) Pixel-based tree species classification: The random forest classifier (RF) was used to classify urban tree species with features derived from LiDAR and DAP data separately. (5) Individual tree species classification: Based on the segmented individual tree boundaries and pixel-based classification results, the majority filtering method was implemented to obtain the final individual tree species classification results. (6) Fused with hyperspectral data: LiDAR-hyperspectral and DAP-hyperspectral fused data were used to conduct individual tree species classification. (7) Accuracy assessment and comparison: The accuracy of the above results were assessed and compared. The results indicate that LiDAR outperformed DAP in individual tree segmentation (F-score 0.83 vs. 0.79), while DAP achieved higher pixel-level classification accuracy (73.83% vs. 57.32%) due to spectral-textural features. Fusion with hyperspectral data narrowed the gap, with LiDAR reaching 95.98% accuracy in individual tree classification. Our findings suggest that DAP offers a cost-effective alternative for urban forest management, balancing accuracy and operational costs.


  • Open Access

    A Down-Scaling Inversion Strategy for Retrieving Canopy Water Content from Satellite Hyperspectral Imagery

    by Meihong Fang, Xiangyan Hu, Jing M. Chen, Xueshiyi Zhao, Xuguang Tang, Haijian Liu, Mingzhu Xu, Weimin Ju

    Vegetation canopy water content (CWC) crucially affects stomatal conductance and photosynthesis and, consequently, is a key state variable in advanced ecosystem models. Remote sensing has been shown to be an effective tool for retrieving CWCs. However, the retrieval of the CWC from satellite remote sensing data is affected by the vegetation canopy structure and soil background. This study proposes a methodology that combines a modified spectral down-scaling model with a high-universality leaf water content inversion model to retrieve the CWC through constraining the impacts of canopy structure and soil background on CWC retrieval. First, canopy spectra acquired by satellite sensors were down-scaled to leaf reflectance spectra according to the probabilities of viewing the sunlit foliage (PT) and background (PG) and the estimated spectral multiple scattering factor (M). Then, leaf water content, or equivalent water thickness (EWT), was obtained from the down-scaled leaf reflectance spectra via a leaf-scale EWT inversion model calibrated with PROSPECT simulation data. Finally, the CWC was calculated as the product of the estimated leaf EWT and canopy leaf area index. Validation of this coupled model was performed using satellite-ground synchronous observation data across various vegetation types within the study area, affirming the model’s broad applicability. Results indicate that the modified spectral down-scaling model accurately retrieves leaf reflectance spectra, aligning closely with site-level measured spectra. Compared to the direct inversion approach, which performs poorly with Hyperion satellite images, the down-scale strategy notably excels. Specifically, the Similarity Water Index (SWI)-based canopy EWT coupled model achieved the most precise estimation, with a normalized Root Mean Square Error (nRMSE) of 15.28% and an adjusted R2 of 0.77, surpassing the performance of the best index Shortwave Angle Normalized Index (SANI)-based model (nRMSE = 15.61%, adjusted R2 = 0.52). Given its calibration using simulated data, this coupled model proved to be a potent method for extracting canopy EWT from satellite imagery, suggesting its applicability to retrieve other vegetative biochemical components from satellite data.


  • Open Access

    Weakly Supervised Semantic Segmentation of Point Cloud Scenes using Boundary-based Feature Aggregation

    by Yongwei Miao, Guoxiang Ren, Xudong Zhang, Haijian Liu, Fuchang Liu

    The task of weakly supervised point cloud semantic segmentation has received widespread attention and also has been widely used in autonomous driving, robotics, and modern industries. Due to the high dimensionality and large volume of point cloud data, many technical difficulties appeal to weakly supervised semantic processing, especially segmentation. However, weakly-supervised schemes usually provide only partial labelling information of the underlying point cloud data and thus need to effectively extract local features, geometric information, and contextual relationships only using these limited labelled data for supervised learning. To improve weakly-supervised semantic segmentation, we propose a novel segmentation network through the boundary-based feature aggregation based on a K-NN algorithm with down-sampling operation, and also introduce the smoothness loss and Siamese loss for effective segmentation. The experiments on public datasets demonstrate that our presented segmentation method exceeds most of the existing fully supervised and weakly-supervised methods in terms of mIoU. Specifically, our network has high segmentation accuracy on the labels of objects with similar geometrical structures, such as ceiling, wall, floor, chair and table, reaching 91.2%, 98.8%, 83.3%, 75.3% and 80.2%, respectively. Extensive experiments also illustrate its robustness, effectiveness, and generalization of the proposed weakly supervised segmentation network.

  • A New Hierarchical Multiplication and Spectral Mixing Method for Quantification of Forest Coverage Changes Using Gaofen (GF)-1 Imagery in Zhejiang Province, China

    by Haijian Liu, Zhifeng Yu, C. K. Shum, Qixia Man, Ben Wang

    The forest survey is a prerequisite or critical for ecological conservation. Spectral mixture analysis is an effective method to extract the forest coverage and its changes, however it is mostly applied to hyperspectral image data processing due primarily to the limit of the number of spectral bands. Therefore, a hierarchical multiplication (HM) model based on the hierarchical method and stepwise multiplication algorithm is proposed for forest coverage extraction using multispectral images. Within the process, the hierarchical method reduces the complexity of the problem to satisfy the SMA at each level, and the multiplication method transfers forest abundances among levels for the soft classification. Compared with Zhejiang Forest Bureau Reports, the HM method in this study extracted 98.43% of forests using the 16-m resolution Gaofen-1 (GF-1) wide field of view (WFV) data, which has a 95% correlation coefficient with results obtained by the 2-m resolution panchromatic / multispectral (PMS) data using the support vector machine (SVM) method. Zhejiang’s overall forest coverage was found to exceed 60% in 2019, with a standard deviation of 0.419 pixels. Densely covered forests are primarily distributed in mountains and hills, and have slightly increased from 2014 to 2019, while sparsely covered forests are mostly located in plains, basins, and valleys, and slightly have declined during the past five years. The forest coverage is mainly affected by topography, population, economy, and policies. The experiment indicates the combined use of HM and multispectral data can accurately extract forest cover changes and achieve similar results compared to more sophisticated classifications using higher precision/spectral band (hyperspectral) data.


  • Open Access

    Analysis on the Rationality of Urban Green Space Distribution in Hangzhou City Based on GF-1 Data

    by Danying Zhang, Haijian Liu, Zhifeng Yu

    With its ecological, economic, and social benefits, urban green space (UGS) plays an important role in urban planning. Accordingly, it is also an important indicator in the evaluation of urban liveability. However, the extraction and statistical analysis of UGS are difficult because urban land use involves complex types and UGS exhibits fragmented distribution and common vegetation extraction models such as the NDVI model and pixel bipartite model. In addition, there are few studies that analyze UGS in Hangzhou with a pixel decomposition model. Therefore, applying the mixed pixel decomposition model with GF-1 data, the following three objectives were set in this study: (1) analyzing the temporal changes of UGS in Shangcheng District, Hangzhou from 2018 to 2020; (2) analyzing the spatial distribution characteristics of UGS in the six main urban areas of Hangzhou in 2020; (3) analyzing the rationality and influencing factors of UGS distribution in Hangzhou. In Shangcheng District, the overall UGS area increased from 2018 to 2020 due to the increase in forest area rather than grassland area. As for the main built-up area in Hangzhou, medium and high coverage of UGS were primarily observed, with an overall high level of greening and a relatively uniform vegetation cover. Only a few areas showed very low UGS coverage. Some differences were observed among administrative regions under the influence of topography, but the overall coverage is high. At the same time, the distribution of UGS in Hangzhou is closely related to policy guidance, the needs of urban residents, and the requirements of economic development. This research not only can provide a new way to analyze UGS features in Hangzhou but also provides scientific guidance for governments in urban planning.


  • Open Access

    Comparative Study on Recognition Models of Black-Odorous Water in Hangzhou Based on GF-2 Satellite Data

    by Zhifeng Yu, Qiyu Huang, Xiaoxue Peng, Haijian Liu, Qin Ai, Bin Zhou, Xiaohong Yuan, Meihong Fang, Ben Wang

    To improve the ability of remote sensing technology in recognizing black-odorous water bodies in Hangzhou, this study analyzed the typical spectral characteristics of black-odorous water in Hangzhou based on measured spectral data and water quality parameters, including the transparency, dissolved oxygen, oxidation reduction potential, and ammonia nitrogen. The single-band threshold method, the normalized difference black-odorous water index (NDBWI) model, the black-odorous water index (BOI) model, and the color purity on a Commission Internationale de L’Eclairage (CIE) model were compared to analyze the spatial and temporal distribution characteristics of the black-odorous water in Hangzhou. The results showed that: (1) The remote sensing reflectance of black-odorous water was lower than that of ordinary water, the spectral curve was gentle, and the wave peak shifted toward the near-infrared direction in the wavelength range of 650–850 nm; (2) Among the aforementioned models, the normalized and improved normalized black-odorous water index methods had a higher accuracy, reaching 87.5%, and the threshold values for black-odorous water identification were 0.14 and 0.1, respectively; (3) From 2015 to 2018, the quantity of black-odorous water in the main urban area of Hangzhou showed a decreasing trend, and black-odorous water was mainly distributed in the Gongshu District and tended to appear in narrow rivers, densely populated areas, and factory construction sites. This study is expected to be of great practical value for the rapid tracking and monitoring of urban black-odorous water by using remote sensing technology for future work.


  • Open Access

    Developing a Scene-Based Triangulated Irregular Network (TIN) Technique for Individual Tree Crown Reconstruction with LiDAR Data

    by Haijian Liu, Changshan Wu

    LiDAR (Light Detection and Ranging)-based individual tree crown reconstruction is a challenge task due to the variable canopy morphologies and the penetrating properties of LiDAR to tree crown surfaces. Traditional methods, including LiDAR-derived rasterization, low-pass filtering smooth algorithm, and original triangular irregular network (TIN) model, have difficulties in balancing morphological accuracy and model smoothness. To address this issue, a scene-based TIN was generated with three steps based on the local scene principle. First, local Delaunay triangles were formed through connecting neighboring point sets. Second, key control points within each local Delaunay triangle, including steeple, inverted tip, ridge, saddle, and horseshoe shape control points, were extracted by analyzing multiple local scenes. These key points were derived to determine the fluctuations of forest canopies. Third, the scene-based TIN model was generated using the control points as nodes. Visual analysis indicates the new model can accurately reconstruct different canopy shapes with a relatively smooth surface, and statistical analysis of individual trees confirms that the overall error of the new model is smaller than others. Especially, the scene-based TIN derived raster reduced the average error to 0.18 m, with a standard deviation of 0.41, while the average errors of LiDAR-derived raster, low-pass filtered smooth raster, and original TIN derived raster have average errors of 0.96, 2.05, and 1.00 m, respectively. The local scene-based control point extraction also reduces data storage due to the elimination of redundant points, and furthermore the different point densities on different objects are beneficial for canopy segmentation.