Table of Content

25 March 2025, Volume 0 Issue 3

Previous Issue   

Deformation extraction of Jinchuan mining area using DS-InSAR with Sentinel-1A ascending and descending orbit

GUO Jie, ZHANG Gonghai, SONG Yewei, BAI Yuxing, HU Jiyuan, WANG Jie, WU Wenhao, GE Zixuan

2025, 0(3):  1-7.  doi:10.13474/j.cnki.11-2246.2025.0301

Asbtract ( )   HTML ( )   PDF (3786KB) ( )  

References | Related Articles | Metrics

The Jinchuan mining area is extremely rich in nickel resources and is the largest nickel metal origin in China. The traditional PS-InSAR method has many problems such as low target point density, deformation area unwrapping errors, and low extraction accuracy of deformation results in the Jinchuan mining area. This paper adopts the advanced DS-InSAR method to process three groups of Sentinel-1A data from different orbits(ascending orbit Path 128, descending orbit Path 33, and descending orbit Path 135) from August 2021 to June 2022. The results are compared with the traditional PS-InSAR method in terms of monitoring point density. The deformation results extracted from the two types of time-series InSAR are cross-validated against the GNSS sites monitoring results. The results show that: ①The MPS density for the mining body monitoring in the Jinchuan mining area is at least 5 times higher than the PS-InSAR method. ②The time-series deformation results solved by DS-InSAR and PS-InSAR are compared with the GNSS sites time-series deformation results in terms of mean absolute error, root mean square error, and Pearson correlation. The three groups of indicators quantitatively reflect that DS-InSAR has the better performance and indicates the deformation results extracted by DS-InSAR are more accurate.

3D deformation prediction of mine surface based on combined SMA-CNN-GRU-Attention modeling

PENG Yibo, YANG Weifang, YAN Xiangrong, GAO Motong, HOU Yuhao, ZHANG Delong

2025, 0(3):  8-14,20.  doi:10.13474/j.cnki.11-2246.2025.0302

Asbtract ( )   PDF (6421KB) ( )  

References | Related Articles | Metrics

Research on monitoring and prediction of surface deformation in mining areas is of great significance for safe production and disaster prevention and warning in mining areas. Existing studies tend to monitor and predict the vertical subsidence of the ground surface, and there are fewer studies on the prediction of 3D directional deformation. To address the above problems, this paper is based on the small baselines set synthetic aperture radar interferometry (SBAS-InSAR) technology to monitor the surface deformation of the west second mining area of Jinchuan mining area with multi-track data. A combined SMA-CNN-GRU-Attention network model with slime mould Algorithm (SMA) is proposed to predict the surface deformation in this area. The results show that adding SMA for optimal parameter solving reduces the MAE and RMSE of the vertical prediction results by 30% and 46% compared to the CNN-GRU network model; the MAE and RMSE of the north-south prediction results are 37% and 39% lower, respectively; and the accuracy of the east-west prediction results is lower, with the MAE and RMSE lower than those of the CNN-GRU network model by 6% and 10%, respectively. The SMA algorithm can accelerate the efficiency of the optimal parameter selection of the model, and it can also improve the prediction performance of the CNN-GRU-Attention model to a larger extent.The SMA-CNN-GRU-Attention multi-feature input prediction model has the superiority compared with other prediction models, and it provides an effective method for the research of 3D deformation prediction of the ground surface.

Application of time series InSAR technology in surface subsidence monitoring and spatio-temporal evolution analysis in mining area

ZHANG Yuxin, YUAN Xiping, GAN Shu, PENG Xiang, WANG Song

2025, 0(3):  15-20.  doi:10.13474/j.cnki.11-2246.2025.0303

Asbtract ( )   PDF (3246KB) ( )  

References | Related Articles | Metrics

In view of the hazards caused by surface subsidence to the safety, environment, socio-economic development and sustainability of resource utilization in the mining area, 63 Sentinel-1A data from December 31, 2021 to March 2, 2024 were first obtained, and SBAS-InSAR (time series interferometry) technology was adopted to monitor the surface deformation of Baicao mining area. The results of surface settlement rate and cumulative settlement in the mine area are obtained, and then the reliability analysis of the monitoring results is carried out by using the measured data. Finally, the settlement of the mine area is predicted based on the LSTM model, and the spatiotemporal variation characteristics and evolution rules of the settlement of the mine area are analyzed in detail.The final conclusions are as follows: ① Spatially, the surface subsidence of Baichuang Mining area is mainly concentrated in the west of the mining area, with the maximum subsidence of -316.86mm and the maximum annual average subsidence rate of -148.4mm/a, and the total subsidence area of 0.6236km², of which the heavy and extremely heavy subsidence area of 0.2804km² needs to be monitored.②In time series, the area with severe subsidence starts to settle from the monitoring starting point, and the subsidence rate tends to be uniform. If no protection is taken, the area will continue to settle in the future, and the settlement may be intensified.③The fitting degree of measured data and monitoring data is high, and the coefficient of determination R² is up to 0.994. The prediction effect of LSTM prediction model on monitoring data is good, and the linear fitting coefficient of determination R² of predicted value and monitoring value can reach more than 0.946, indicating that the prediction of surface settlement by LSTM model can meet the requirement of accuracy. The experimental results can provide technical support for disaster prevention and control in mining areas, and provide strong support for more accurate surface deformation evaluation in mining areas.

Dynamic monitoring for open-pit mine reclamation based on UAV oblique photogrammetry

ZHONG Weihua, LIU Jingkuang

2025, 0(3):  21-26.  doi:10.13474/j.cnki.11-2246.2025.0304

Asbtract ( )   PDF (8280KB) ( )  

References | Related Articles | Metrics

Traditional open-pit mine reclamation monitoring relies on satellite remote sensing and on-site investigation, but suffers from low precision and efficiency. To enhance dynamic monitoring and improve oversight, this paper explores UAV oblique photogrammetry. Using an open-pit mine in Guangzhou as a test case, the technology generated a real-life 3D Mesh model across four phases, synchronously outputting DEM and producing DLG through stereo acquisition. Monitoring indices such as earth backfill elevation, volume, and building demolition were analyzed using DEM and DLG data. Results indicate that UAV data collection over 1.56km² took about one hour, three minutes, and fifty-five seconds, with a GSD of 2.51cm/pixels, marking an improvement in efficiency and accuracy over traditional methods. Additionally, the DLG data helped count a demolition area of 16442.36m², and DEM data allowed for the calculation of a total earth backfill volume of 0.017km³, demonstrating the technology's digital and visual analysis capabilities. This study offers technical support for regional authorities in dynamically monitoring open-pit mine reclamation and provides references for monitoring other mines.

Analysis of spatio-temporal variations and drivers of habitat quality in the coal mine area of Yangquan city

ZHANG Nan, CHEN Shenghua, SUN Caixia

2025, 0(3):  27-32.  doi:10.13474/j.cnki.11-2246.2025.0305

Asbtract ( )   PDF (4075KB) ( )  

References | Related Articles | Metrics

The impact of coal mining on the habitat quality of coal mining area in Yangquan city is solved.Based on Landsat images from 2003 to 2023, this paper classifies the land use types in Yangquan coal mining area, uses the InVEST model to analyze the temporal and spatial evolution characteristics of habitat quality in coal mining area, and uses geographic detectors to explore the driving factors affecting the spatial distribution of habitat quality.The results show that:① In the past 20 years, the land use types are mainly composed of forest land and grassland, the cultivated land area is generally stable, and the construction land has shown an increasing trend year by year.② The habitat quality grade was mainly composed of medium and high grades, but the changes mainly changed from high grade to low grade, and the area area in a state of deterioration accounted for nearly 70%, and the overall habitat quality status deteriorated.③ Land use is the main driver of differences in the spatial distribution of habitat quality,and the interaction of the other influences with land use had the strongest explanatory power.The results of the study provide a theoretical basis for ecological environment management and restoration of government departments.

3D small object detection method based on image and point cloud fusion

HAO Jia, YAO Guoying, ZHOU Jian, WANG Siyuan, XIAO Jinsheng

2025, 0(3):  33-38.  doi:10.13474/j.cnki.11-2246.2025.0306

Asbtract ( )   HTML ( )   PDF (3828KB) ( )  

References | Related Articles | Metrics

Object detection technology plays a pivotal role in key fields such as artificial intelligence, facial recognition, and autonomous driving. 3D point cloud object detection, especially for small objects, remains a significant challenge in technological development. To address this challenge, this paper proposes a novel 3D detection network that integrates image and point cloud data to significantly enhance the accuracy of 3D small object detection. The approach begins by leveraging Yolov5 for precise 2D object detection and establishing a 3D constraint using the coordinate mapping relationship between cameras and LiDAR to extract conical regions of interest from the raw point cloud data. Furthermore, to tackle the issue of detecting small objects in distant point clouds, a cluster-optimized 3D detection network architecture is introduced. This architecture simultaneously inputs the point clouds of the regions of interest into both the PointNet and clustering modules, and then fuses their detection results to improve the accuracy of 3D small object detection. Testing on the KITTI dataset shows that, compared to existing techniques, the proposed algorithm improves the average precision (AP) for two small object categories by 15.94% and 2.29% under moderate difficulty conditions, and by 13.34% and 2.86% under high difficulty conditions. These results underscore the significant impact and practical application potential of this algorithm in enhancing the accuracy of 3D small object detection.

Fusion of time-series InSAR and ESMD for potential hazard identification and deformation monitoring of high-fill expansive soil airports

ZHANG Shuangcheng, LI Sijiezi, REN Zhipeng, SI Jinzhao, HU Xingqun

2025, 0(3):  39-45.  doi:10.13474/j.cnki.11-2246.2025.0307

Asbtract ( )   PDF (19201KB) ( )  

References | Related Articles | Metrics

In this paper,the SBAS-InSAR technique is firstly applied to identify the deformation hazardous area of Ankang expanded soil airport,and obtain the long time series of the characteristic points of the hazardous area.Then,based on the ESMD algorithm,the deformation time series is decomposed,the seasonal physical signals are extracted,and the surface deformation signals are highlighted to obtain the seasonal band cycle of the time series and the frequency and time of the occurrence of the deformation anomalies.Finally,the deformation factors are analyzed in combination with the rainfall and temperature factors in the environmental loads. The results show that:①The deformation of the airport surface mainly occurs in the area of expansive soil fill,and the deformation of the high-fill slopes of the airport is especially obvious. ②The deformation of expansive soils is subject to seasonal fluctuations due to environmental factors and is prone to inhomogeneous settlement in summer. The results of this study provide important clues about the mechanism of cyclic deformation of expansive soils,and are useful for the monitoring of transportation infrastructures under expansive soil conditions.

A method of extracting streetlights based on vehicle-borne laser point clouds

ZHANG Fujie, WANG Liuzhao, ZHONG Ruofei, XU Mengbing, JIN Huanhuan

2025, 0(3):  46-51.  doi:10.13474/j.cnki.11-2246.2025.0308

Asbtract ( )   PDF (3646KB) ( )  

References | Related Articles | Metrics

Streetlights are critical components of urban infrastructure,timely and accurate acquisition of streetlight information is essential for the development of digital cities. Constrained by the complex object structures and occlusions in urban environments,traditional streetlight extraction methods still suffer from low accuracy,inefficiency,and poor robustness. Additionally,these methods lack general applicability across different urban scenarios. To address these issues,this paper proposes an automatic method for extracting urban streetlights based on vehicle-borne laser point clouds. Firstly,a cylindrical spatial neighborhood is established using ISS keypoints,and potential pole-like objects are identified through density threshold discrimination and back-projection. Then,non-target pole-like objects,such as street trees,are rapidly eliminated using PCA principal vectors,normal vector directions,and angular constraints,resulting in a candidate set of streetlight points. Finally,leveraging the spatial geometric features of streetlight point clouds,a decision tree model is instantiated via a random forest algorithm to match and classify the candidate streetlights,achieving precise extraction of streetlight point clouds. Experimental results indicate that the proposed method attains high extraction accuracy and robustness when dealing with regularly distributed or partially occluded streetlight point clouds,demonstrating significant practical application value.

DGIONet:dual-path global information optimization network for sea-land segmentation with remote sensing images

XIE Batu, HU Jiarui, PAN Jun

2025, 0(3):  52-58,86.  doi:10.13474/j.cnki.11-2246.2025.0309

Asbtract ( )   HTML ( )   PDF (3440KB) ( )  

References | Related Articles | Metrics

Addressing the impact of coastal features on the refined segmentation of land-sea boundaries in high-resolution remote sensing images, this paper proposes a land-sea segmentation algorithm for remote sensing images based on DGIONet. In the encoding stage, the network is designed with a multi-scale spatial attention feature extraction module based on rectangular strip convolution, which utilizes rectangular strip convolutions constructed vertically at different scales to achieve a multi-scale large kernel convolution effect. Relying on the extracted multi-scale features and point convolutions within the module, it implements a spatial attention mechanism, effectively enhancing the network's ability to focus on large scale land-sea features, and enabling feature extraction of land-sea global information and contextual information. In the decoding stage, the network incorporates a dual-path global information optimization decoder. Within the decoder, it relies on a depthwise separable dilated convolution information optimization module and a “Hamburger” global feature restoration module to leverage the extracted global and contextual information for feature restoration. Meanwhile, it fuses the stage features extracted from the encoding stage with the stage-restored features to achieve better information optimization. To verify the effectiveness of the proposed method, this paper constructs a high-resolution remote sensing image land-sea segmentation dataset with a resolution better than 0.3m. Comparative experiments conducted on this dataset demonstrate that the proposed method improves pixel accuracy by 3.01% and mean intersection over union(mIoU) by 10.51% compared to the current mainstream semantic segmentation method, Vision Transformer. This proves the advantages of the proposed method in the refined land-sea segmentation of high-resolution remote sensing images.

Precise determination of bolt holes in multiple source shield tunnels by multi-feature template matching

WANG Shaoning, YANG Yuanwei, XU Lei, GU Shicheng, GAO Xianjun, YIN Zhenghao, ZHONG Kang, LIU Zhenyu

2025, 0(3):  59-65.  doi:10.13474/j.cnki.11-2246.2025.0310

Asbtract ( )   PDF (4505KB) ( )  

References | Related Articles | Metrics

Given that the bolt hole is the primary mechanical component of the shield tunnel,its overall structural stability depends heavily on it.A possible risk to the secure functioning of tunnels is disease phenomena such water seepage and cracks brought on by bolt failure.In order to eliminate the potential safety hazards of inspectors and improve the detection efficiency of bolt holes,this paper proposes an accurate identification method for bolt holes in multi-source shield tunnels based on multi-feature template matching.Firstly,the geometric center of the tunnel section point cloud is chosen as the viewpoint,and the tunnel point cloud is expanded into a 2.5 D point cloud using cylindrical projection.The bolt hole point cloud is then extracted using DBSCAN,and its center coordinate set is generated using Euclidean clustering.The DBSCAN clustering method based on the shape features of the point cloud designed in this paper can extract all bolt holes from the point cloud data.The average similarity and recognition rate of the proposed method can reach 98.79% and 98.76%,respectively,and the average deviation is smaller and more robust in the case of similar time.In this paper,the shape characteristics of the bolt hole point cloud are fully considered to realize the accurate classification of the bolt hole point cloud,and the accuracy of target recognition on the shield tunnel image is further improved by fusing the three-dimensional and two-dimensional data of the shield tunnel.

Water extraction from Sentinel-1 images based on improved DeepLabV3+ network

ZHAO Xingwang, ZHAO Yan, LIU Chao, LIU Chunyang

2025, 0(3):  66-70.  doi:10.13474/j.cnki.11-2246.2025.0311

Asbtract ( )   PDF (2008KB) ( )  

References | Related Articles | Metrics

In order to improve the accuracy of water extraction from radar images, this paper uses Sentinel-1 series images from 2023 as the data source, optimizes the backbone network on the basis of the DeepLabV3+ network model, integrates the SE channel attention mechanism, and proposes an improved deep learning network model SEDeepLabV3+. The ablation experiment is carried out for the improved model, and the model is verified by the water body extraction in Changping district of Beijing on July 31. The experimental results show that when the improved SEDeepLabV3+ method is used to extract water body, the mean intersection over union and pixel accuracy can reach 88.55% and 93.49%. Compared with DeepLabV3+, HRNet and U-Net, the average intersection ratio is increased by 2.26%,2.31% and 5.08%,and the average pixel accuracy is increased by 0.76%,0.80% and 3.07%,respectively. The improved SEDeepLabV3+ not only has a lighter network structure,but also can effectively improve the accuracy and efficiency of water extraction.

Identification of landslide hazards of Beipanjiang Guangzhao hydropower station based on Lutan-1 SAR satellite

XU Qingsong, HU Jun, CUI Wengang, LIU Suihua, HU Dan, NING Fei, LI Man

2025, 0(3):  71-75,132.  doi:10.13474/j.cnki.11-2246.2025.0312

Asbtract ( )   PDF (21550KB) ( )  

References | Related Articles | Metrics

Lutan-1 (LT-1) SAR satellite is China's first L-band fully polarimetric civil interferometric SAR satellite constellation.It is composed of two SAR satellites with the same parameters,and the revisit period can reach 4 days in the follow-up mode,which has the advantages of high resolution,short revisit period and full polarization imaging.It is conducive to the identification of landslide hazards in complex areas. In this paper,13 scenes of LT-1 ascending SAR data from April 2,2023 to June 21,2023 are obtained,and the differential interferometry synthetic aperture radar (D-InSAR) is used to identify landslide hazards,and the reliability of the identification results is verified by combining optical remote sensing images. The results show that there are 6 landslide hazard areas in Beipanjiang Guangzhao hydropower station,which are mainly distributed in the north-east direction of the reservoir area,and the spatio-temporal evolution characteristics of the two typical hazard areas are analyzed in detail through comprehensive rainfall and optical remote sensing images,and it is found that rainfall is one of the factors affecting slope stability,which proves the feasibility and effectiveness of LT-1 satellite in landslide hazard identification,and provides a new data source for the geological disaster prevention and control industry.

3D model construction and optimization based on vertical aerial photography

LIU Jincang, YUAN Lei, MA Defu, XIA Jinliang

2025, 0(3):  76-80.  doi:10.13474/j.cnki.11-2246.2025.0313

Asbtract ( )   PDF (6297KB) ( )  

References | Related Articles | Metrics

Vertical aerial photography is generally used to obtain image data to make orthographic images, in order to further digging the value of image data and support the construction of China's national 3D mapping program, this paper studies the construction of three-dimensional model based on the image obtained by vertical aerial photography and explores the method of model effect optimization, specifically include the analysis of the feasibility of constructing a real 3D model based on vertical aerial images, the study of the method of enhancing the effect of the three-dimensional model by increasing the image overlap without increasing the cost of aerial photography, and the proposed model optimization method for building top information constraints. Through theoretical analysis and practice, it is proved that: the overlap has a great influence on the integrity of the three-dimensional model based on vertical aerial images, using 80 % overlap degree of heading to obtain image data can greatly improve the modeling effect without increasing the workload of aerial photography; Building top information constraint modeling method is very effective to improve the integrity of the model,The modeling method of building top information constraint is very effective to improve the integrity of the model; Building top information constraint modeling method is very effective to improve the integrity of the model, it can greatly optimize the edge shape of the building under the premise of retaining the detailed information of the top of the building, and can play an important role in urban planning, law enforcement monitoring and other fields.

Multi-point hemispherical grid model with adaptive grid division for BDS multi-path errors based on AMR clustering

WANG Yawei, WU Baifa, HUANG Lei, LIU Huaguang, WU Zhiwen, ZHAN Yanchun, LI Haiyang

2025, 0(3):  81-86.  doi:10.13474/j.cnki.11-2246.2025.0314

Asbtract ( )   PDF (3806KB) ( )  

References | Related Articles | Metrics

In response to the computational resource consumption challenge associated with fixed-resolution multi-point hemispherical grid model MHGM. A method for adaptive grid division of the station's hemisphere using spatial domain prior distribution information based on multi-path errors in the BDS is proposed. The experimental results indicate that this method can achieve adaptive division of the grids in MHGM at the station, merging areas with weak multi-path errors and appropriately densifying areas significantly affected by multi-path errors. As the adaptive threshold parameter k increases, the correction effect of the MHGM on the multi-path errors of BDS shows a slight decrease compared to the fixed resolution, but it is still significantly better than the correction effect of ESM used as prior information. When the value of k is set to 0.9cm, compared to the fixed resolution mode, the number of parameters to be estimated in the MHGM after adaptive grid division in the experiment decreases by 77.2%, while the memory usage during parameter estimation is only 5.2% of that in the fixed resolution. This proposed method, while ensuring the effectiveness of multi-path error reduction for the BDS, further highlights its potential value in multi-path errors modeling for GNSS large-scale network data.

A new towed water depth measurement system

SHEN Wei, WANG Zhangjiangyao, WANG Zicheng, DING Zihan, FENG Qibin

2025, 0(3):  87-92.  doi:10.13474/j.cnki.11-2246.2025.0315

Asbtract ( )   HTML ( )   PDF (5223KB) ( )  

References | Related Articles | Metrics

Aiming at the urgent requirements of the current large-scale water survey and bathymetric work, this paper designs a new towed bathymetric survey system. The system makes creative improvements to the existing single-beam bathymetry device, and designs a modular electronic warehouse and floating body, which can be quickly deployed and recovered by unmanned aerial vehicles and other equipment, and carry out bathymetry survey by towing on the surface of the water.CFD simulation analysis and field tests show that the system exhibits good resistance and stability under different water velocities, with a drag force of 53.36N for the system when the maximum water velocity is 3m/s. The internal conformity accuracy of the system for bathymetry in open water is 0.0245m, and that of the unmanned ship is 0.0342m. The experimental accuracy is in accordance with the requirements of the relevant national specifications. This system with reliable accuracy and stable operation,can be quickly deployed and recovered,and is especially suitable for measurements in waters with steep slopes where unmanned boats can not be deployed, or in waters such as densely arranged pits and ponds. The results validate the application potential of the proposed proposed system in complex waters and provide an innovative solution for bathymetric surveying.

Adaptive map matching algorithm using second-order hidden Markov models for complex scenarios

GUO Siyu, GUO Yuan, LI Bijun, WU Chaozhong

2025, 0(3):  93-98.  doi:10.13474/j.cnki.11-2246.2025.0316

Asbtract ( )   HTML ( )   PDF (2772KB) ( )  

References | Related Articles | Metrics

As the complexity of urban transportation systems has significantly increased, existing map matching methods still face considerable challenges in handling complex urban traffic scenarios such as intersections and overpass obstructions. To address these issues, we propose a map-matching method tailored for complex urban road networks.Firstly, through the features of directionality and connectivity, we quantify the complexity of the road network scene where trajectory points reside and achieve trajectory segmentation. Then, for simple trajectories, we use a direction-constrained hidden Markov model (HMM) for matching, while for complex trajectory segments, a second-order model is adopted that uses the complexity of the road network as a weighting parameter to adaptively adjust the ratio of observation probabilities and transition probabilities in the HMM, improving the accuracy and efficiency of map matching in complex road networks.Finally, we compare traditional HMM methods with the ST-Matching method.The results show that the proposed algorithm improves matching accuracy by 5.4% and 6.0% respectively in complex scenarios and has higher matching efficiency.

Long-term poverty level estimation based on multi-source geographic data: a case study in Bangladesh

JIANG Ming, ZHANG Fuhao, ZHAO Xizhi, ER Geli, YU Hao

2025, 0(3):  99-104.  doi:10.13474/j.cnki.11-2246.2025.0317

Asbtract ( )   PDF (2543KB) ( )  

References | Related Articles | Metrics

Poverty is a major social issue commonly faced by developing countries. To address the problem of long-term poverty data gaps, this study proposes a method for constructing a comparable wealth index (CWI) based on household survey data, which is used to represent poverty level. On this basis, a poverty estimation method is proposed,which integrates multi-source spatiotemporal features and a random forest algorithm. The method utilizes data such as nighttime light remote sensing, roads, land cover, digital elevation models, and flood inundation zones to estimate poverty levels in Bangladesh from 2014 to 2021. Experimental results show that the proposed CWI construction method and poverty estimation approach are feasible, with an R² value of 0.88 for the poverty estimation model, indicating high accuracy. The findings can serve as a reference for poverty estimation and analysis in other developing countries.

Deep learning-based disease detection framework for ultra-high resolution images of tunnels

MA Haizhi

2025, 0(3):  105-110.  doi:10.13474/j.cnki.11-2246.2025.0318

Asbtract ( )   PDF (4910KB) ( )  

References | Related Articles | Metrics

The data collected by existing tunnel detection techniques usually obtains ultra-high resolution images, and the actual area of the disease in the tunnel is small, which makes the loss of disease information occur after the image has been simply pre-processed (e.g., scaled), and the deep learning model trained under limited computational resources may have a reduced detection rate of the object, unstable training, and other phenomena. To address the above problems, this paper proposes a framework for disease detection based on deep learning and ultra-high resolution images of tunnels, which is applicable to any deep learning model by pre-processing the ultra-high resolution image, segmenting the original image into smaller patch images, and resizing the ultra-high resolution image to a suitable size to improve the performance of the detection model. The experimental results show that the performance of the model under the proposed framework improve by about 77.19% compared with conventional detection process. And the proposed framework is applicable to general ultra-high resolution images, which can effectively identify the damages of general structures other than tunnels.

Concrete crack detection algorithm based on super-resolution generative adversarial network

LI Xiang

2025, 0(3):  111-116.  doi:10.13474/j.cnki.11-2246.2025.0319

Asbtract ( )   PDF (2678KB) ( )  

References | Related Articles | Metrics

With the growth of service life, tunnels will inevitably undergo aging, and as an important infrastructure for the travelling of urban residents, tunnel safety inspection is crucial. At present, the crack disease on the tunnel surface is mostly detected using the images taken by cameras. However, the cracks have a small pixel percentage in the image, and its detection process is time-consuming and labor-intensive. Hence, there is an urgent need for a method that can accurately detect the cracks in a large field-of-view range. This paper proposes a learning structure based on super-resolution generative adversarial networks, which is applicable to any segmentation network, and proposes a method for efficiently constructing training data to be applied to the proposed learning structure. The performance of the proposed method is evaluated on 1606 crack images with randomly degraded quality. The results show that the crack detection IoU and F1 scores under the proposed learning structure are 63.686% and 77.811%, respectively, and the variances are 0.9008 and 0.5015, which effectively improves the performance of crack detection and has high robustness to the input data.

Disease development analysis of cross-river shield tunnel based on SHAP explanation of random forest

XU Pengyu, WANG Yong, WANG Hongxue, GAO Xiang

2025, 0(3):  117-121.  doi:10.13474/j.cnki.11-2246.2025.0320

Asbtract ( )   PDF (1740KB) ( )  

References | Related Articles | Metrics

Rail transit construction technology in China is becoming more and more mature, however, the current situation of tunnel deformation in a water-rich environment can not be changed. In the field of engineering, scholars have conducted in-depth research on the causes, influencing factors, control and treatment of tunnel diseases. However, there are few studies on the disease distribution and development prediction of tunnels, especially cross-river shield tunnels. In order to make up for the blank in the field of disease development prediction of cross-river shield tunnels, the study forecasts the next phase of monitoring data based on RF added SHAP values and historical monitoring data which are from the full-section scanning of the shield tunnel in the cross-river section by 3D laser measurement technology. After evaluating the accuracy of the prediction results, we can analyze the prediction data to determine the location and degree of the tunnel deformation, which will provide the basis for subway operation and maintenance.

A tunnel lining health evaluation method with ground-penetrating radar and deep learning

ZHANG Guangwei

2025, 0(3):  122-126,149.  doi:10.13474/j.cnki.11-2246.2025.0321

Asbtract ( )   PDF (2111KB) ( )  

References | Related Articles | Metrics

Tunnel in its service life, affected by a variety of factors, behind the tunnel wall will produce a variety of structural diseases such as voids, incompact, and so on, affecting the service performance, the ground-penetrating radar (GPR) nondestructive testing technology is widely used in the field of tunnel quality inspection, but due to the complexity of the radar data deciphering work and the large amount of data, the detection efficiency needs to be improved. In recent years, machine learning has attracted much attention due to its excellent data processing capability and information extraction ability, providing a variety of efficient and reliable disease classification models. In this paper, based on ground-penetrating radar images, a multilevel disease classification method is proposed for assessing tunnel lining health. Firstly, in this paper, radar image data are acquired and manually decoded to create a sample database to be used as inputs and outputs of the model in order to train and test the deep learning model. For the small sample characteristics of the database, the data are classified using the Vision Transformer network and the improved Compact Convolutional Transformer. The results show that the Vision Transformer algorithm can achieve tunnel lining health evaluation based on radar images with better results and high accuracy compared to other versions.

Refined extraction of multi-feature inland water bodies in Zhejiang province

WANG Xingkun, LI Jiaxin, FENG Cunjun, ZHAN Yuanzeng, ZHU Xiaojuan, ZHOU Wei, DENG Xiaoyuan

2025, 0(3):  127-132.  doi:10.13474/j.cnki.11-2246.2025.0322

Asbtract ( )   PDF (5437KB) ( )  

References | Related Articles | Metrics

Aiming at the problem of insufficient accuracy of automatic extraction of multi-feature inland water bodies by satellite remote sensing,this paper takes Zhejiang province as the research scope to explore the extraction accuracy of the Vision Transformer (ViT)largevision model for inland water bodies with different features. Through the results of historical geographical national conditions monitoring,large scale samples are obtained to get a pre-trained model.Combined with the multi-level perception characteristics of inland water bodies in Zhejiang province,the UPerNet network is used to optimize the output layer of the ViT model from different aspects such as scenes,objects,parts,materials and textures,further increasing the perception ability of the ViT model for multi-scale and multi-feature water bodies. The accuracy and recall rate of the algorithm in this paper can reach 90%,which is 15% higher than the traditional exponential threshold method and 10% higher than the pre-trained model.Through post-processing,it can meet the accuracy requirements of the water surface area survey and monitoring in Zhejiang province.The results show that the feature-optimized visual large model can be well applied to the extraction of multi-feature inland water bodies and serve the national water resources survey.

Improved DeepLabV3+ model for spring crops identification in Sichuan counties

ZHANG Xuan, YANG Benyong, WEN Wu, DENG Weixi, ZHOU Hefan

2025, 0(3):  133-137,167.  doi:10.13474/j.cnki.11-2246.2025.0323

Asbtract ( )   PDF (7915KB) ( )  

References | Related Articles | Metrics

Traditional agricultural irrigation water resource allocation models suffer from significant inefficiencies and waste due to uneven distribution. Remote sensing technology can effectively address the issue of missing crop distribution data in irrigation planning by providing accurate spatial distribution information. This paper takes Xinjin district of Chengdu, Sichuan province as the research area. A mechanism which contains contrastive learning and feature enhancement is introduced to improve the DeepLabV3+ model, utilizing GF series satellite imagery to accurately identify both major and minor crops. The results indicate that the improved IM-DeepLabV3+ model can achieve recognition accuracies of 91.73%, 89.93%, 80.18%, and 72.08% for rapeseed, wheat, rice, and corn, respectively, which can provide scientific crop distribution data support for scientific allocation of agricultural irrigation water resources.

The extraction method of tea plantation information based on multi-dimensional features from UAV remote sensing images

YANG Jiafang, YIN Linjiang, ZHANG Hongliang, ZHAO Weiquan, LI Wei

2025, 0(3):  138-143.  doi:10.13474/j.cnki.11-2246.2025.0324

Asbtract ( )   PDF (2619KB) ( )  

References | Related Articles | Metrics

As a crucial economic crop, the rapid measurement of tea planting areas holds significant value for estimating tea yield, optimizing tea garden management strategies. The study employs UAV to obtain multispectral remote sensing images of tea plants. It selects vegetation indices, texture features, and their combined features as multivariate analysis indicators. By using the RF model and correlation analysis, the study evaluates the importance of these indicators and performs correlation tests. Subsequently, three supervised classification algorithms, namely MLC, SVM, and RF are utilized to precisely identify the distribution of tea plants plantations in the study area. The results reveal that: ①When using the SVM and RF algorithms, the incorporation of texture features or multi-feature combinations notably enhances classification accuracy compared with using a single vegetation index feature. ②The optimal classification feature combination for the study area is identified as the visible vegetation indices combined with texture features, achieving a total classification accuracy of 95.5% and a Kappa coefficient of 0.917 with the SVM algorithm. ③The use of full-feature datasets and feature dimensionality reduction datasets does not significantly enhance classification accuracy; however, the latter demonstrates the highest stability in classification results and achieves an accuracy that is second only to the optimal classification feature datasets. In summary, by utilizing the SVM algorithm,the fusion of visible vegetation indices and texture features can effectively differentiate tea plants from other land features,achieve high-precision extraction of tea planting information,and provide valuable reference and practical guidance for accurately extracting crop planting information.

Application of UAV photogrammetry in the survey of sandy coast ecosystem

YU Wei, LIU Mincong, WANG Meng, LIU Yue, LI Jikun, YANG Jie, LI Tuanjie

2025, 0(3):  144-149.  doi:10.13474/j.cnki.11-2246.2025.0325

Asbtract ( )   PDF (4823KB) ( )  

References | Related Articles | Metrics

The sandy coast is a very high-quality tourism resource with high tourism leisure value and aesthetic value. The plane accuracy of using drones to generate orthophoto images of sandy coasts reaches the centimeter level, and the elevation accuracy of digital elevation models reaches the sub meter level. Based on drone orthophoto images, the green leaf index(GLI), green red vegetation index(GRVI), and red green blue vegetation index(RGBVI) are applied to identify the vegetation area of backshore. Compared with the results of GIS digitization, the errors are 24.4%, 7.4%, and 25.2%, respectively. The results of green red vegetation index(GRVI) identification are closest to those of GIS digitization. There are 66 types of vegetation on the backshore of Dameisha beach, most of which are garden and green vegetation, with fewer native vegetation. The beach area is 16.3 hectares, of which the dry beach area is 8.1 hectares and the intertidal zone area is 8.2 hectares. The drone is not connected to the RTK signal, and the average difference between the elevation data extracted from the generated DEM and the ground RTK measured elevation is 0.598 m. The measured elevation of the drone is lower than the RTK measured results. We use a five lens aerial camera to obtain real land information and construct 3D model of the beach, and generate specific 3D model of the beach sculpture and local buildings behind the backshore.

Improved U-Net convolutional network application for land cover classification in remote sensing images

GOU Changlong, PANG Min, YANG Yang

2025, 0(3):  150-155.  doi:10.13474/j.cnki.11-2246.2025.0326

Asbtract ( )   PDF (2122KB) ( )  

References | Related Articles | Metrics

Land cover classification plays a crucial role in environmental monitoring, resource management,and urban planning. However,the classification process faces various challenges due to factors such as spectral similarity,noise interference,and the intermixing of natural and man-made objects. To improve classification accuracy and enhance the robustness of the model,this paper proposes a deep learning network based on the U-Net convolutional architecture,combined with the Transformer self-attention mechanism.Experiments conducted on the remote sensing dataset of Lanzhou city demonstrate that the proposed model outperforms PSPNet,DeeplabV3,Segformer,and Swin-T in terms of average classification accuracy (mAcc),mean intersection over union (mIoU),and mean F1 score (mF1). In addition to improving classification accuracy,the model achieves high inference speed,showcasing its potential for applications in complex land cover scenarios and offering new insights for remote sensing image classification.

Few-shot defect recognition method of underground drainage pipelines

JU Feng, QIAN Qiangqiang, YANG Zhen, YOU Jiajun

2025, 0(3):  156-160.  doi:10.13474/j.cnki.11-2246.2025.0327

Asbtract ( )   PDF (1798KB) ( )  

References | Related Articles | Metrics

The inspection, evaluation and maintenance of urban underground drainage pipelines are the necessary means to ensure the safe of drainage pipeline system. Deep learning technology provides a new method for the automation and intelligence of underground drainage pipelines defects detection and recognition. However, the lack of sample set for some pipeline defect and the imbalance of samples among different defect types greatly affect the generalization ability and robustness of defect recognition models, resulting in the problems of false detection, missed detection, and low recognition accuracy of existing defect recognition models. To solve the above problems, this paper proposes a defect recognition method for few-shot defect based on metric learning. The method uses detailed features to represent the test images, and identifies the types of defects according to the similarity between the feature vectors of different defect detection images in the support set and the images in the query set. The experimental results show that the accuracy of the proposed method in identifying scarce pipeline defect types is about 65%, and can be used as a solution for intelligent recognition in the case of insufficient and unbalanced pipeline defect samples.

A construction waste pile detection and identification method based on improved U-Net algorithm

ZOU Weilin, ZHOU Wen, ZHANG Yongli, GAO Siyan, WANG Puliang

2025, 0(3):  161-167.  doi:10.13474/j.cnki.11-2246.2025.0328

Asbtract ( )   PDF (5599KB) ( )  

References | Related Articles | Metrics

At present, China's actual production and construction of large volumes of construction waste and complex composition, if not properly handled, some of the components will react with the surrounding environment to form a health hazard, resulting in immeasurable consequences. In addition, under the policy of “waste-free city”, construction waste has become an important issue in solving environmental problems. This paper proposes an improved model of U-Net algorithm based on the construction waste dataset collected independently. Relying on the original U-Net network, the model introduces ResNet residual network, wavelet transform and attention mechanism module into the backbone network, which not only effectively solves the problems of gradient disappearance and blurring of edge features of the original model, but also improves the performance indexes such as mIOU, mPA,F1 score, etc., compared with some other models, and the overall performance of the model is smoother, so that the model can successfully and efficiently accomplish the construction waste stacking and disposal. The overall performance of the model is smooth, which can successfully and efficiently accomplish the task of recognizing and detecting the construction waste dumping.Finally, the streets under the jurisdiction of Weidong district, Pingdingshan city, Henan province, are used as the experimental area for the application of the results to verify the results. The experimental results show that the identification and detection model can effectively identify and detect the coverage of construction waste, and the accuracy of identification and detection meets the requirements of practical application, which can provide important decision support for the realization of construction waste management and disposal.

Research and practice on intelligent extraction mode of surveying and mapping geographic information achievements: a case study of Keqiao district in Shaoxing city

XIONG Lan, HUANG Weixiang, HE Longgang, WANG Qiang, LU Guangcan, ZHANG Chang

2025, 0(3):  168-173.  doi:10.13474/j.cnki.11-2246.2025.0329

Asbtract ( )   PDF (2740KB) ( )  

References | Related Articles | Metrics

Drawing upon the research of extant methodologies and in compliance with the mandates of national institutions for establishing a novel security framework for surveying and mapping geographic information, we investigate an intelligent extraction mode tailored for the distribution of surveying and mapping geographic information achievements. Subsequently, an achievements management process system is developed within this framework. This practice is executed in collaboration with the pilot project in Keqiao district, Shaoxing city. The empirical achievements affirm the feasibility of this model and its potential as a valuable reference for diverse achievements distribution tasks. It notably enhances the efficiency and precision of achievements distribution, while also fortifying the security of achievements data in the context of practical applications.

“Air-ground-human” cooperate to build a real 3D model: taking Haitang pavilion as an example

QIAO Yue

2025, 0(3):  174-177.  doi:10.13474/j.cnki.11-2246.2025.0330

Asbtract ( )   PDF (1663KB) ( )  

References | Related Articles | Metrics

To address the problem of data loss in the reconstruction or modeling process of single data source in the salvage digital preservation project, this paper combines multi-source data, taking the Huanxiu villa Haitang pavilion as an example, and adopts the coordination of “space-land-people” to strengthen the collection of spatial information of the Haitang pavilion, including the survey of basic information of the building and the mapping of the building ontology, so as to build a salvage digital preservation project with high-quality live 3D model. The high-quality realistic 3D model of the digital protection project retains a full range of data and information for the ancient architecture. On the one hand,it provides detailed information for the formulation of long-term management mechanism for the protection of ancient architecture, maintains the virtuous cycle of architectural protection and utilization, realizes the revitalization and updating of ancient architecture, and achieves the retention of a full range of ancient architectural and high-precision status quo information. On the other hand, it provides information for the scholars to study the traditional architecture of China, promotes the inheritance of traditional architectural crafts, and provides powerful data support for the restoration and reconstruction of ancient buildings in the future.

Design and implementation of an intelligent yard management platform based on UAV: a case study of Rizhao port

CHENG Jie, ZHONG Yong, YANG Yugang, JIAO Baotong, ZHANG Fusheng

2025, 0(3):  178-182.  doi:10.13474/j.cnki.11-2246.2025.0331

Asbtract ( )   PDF (3985KB) ( )  

References | Related Articles | Metrics

With the continuous growth of global trade, ports, as critical nodes in the logistics chain, have their operational efficiency and management levels directly impacting the effectiveness of the overall logistics system. Rizhao port, Shandong port, an important bulk cargo hub in China, faces challenges such as complex yard management and inefficiency. Addressing the practical needs of Rizhao port, this paper designs and implements an intelligent yard management platform based on unmanned aerial vehicles. It aims to enhance the level of intelligence in yard management by integrating information technology such as artificial intelligence, improves yard utilization through intelligent stacking operations by UAV, and reduces operational costs. This paper elaborates on the design concept, technical architecture, key functions, and innovations of the platform, and conducts an empirical analysis using Rizhao port as a case study. The results show an increase in stacking efficiency by over 90%, significantly enhancing the operational efficiency of yard stacking management. Furthermore, the analysis and assessment of yard storage capacity indicate an improvement of over 10% in the port's storage capacity.