Table of Content

25 December 2023, Volume 0 Issue 12

Previous Issue    Next Issue

Improved multi-task road feature extraction network and weight optimization

ZHU Wenjie, LI Hongwei, JIANG Yirui, CHENG Xianglong, ZHAO Shan

2023, 0(12):  1-7.  doi:10.13474/j.cnki.11-2246.2023.0350

Asbtract ( )   HTML ( )   PDF (4323KB) ( )  

References | Related Articles | Metrics

In order to address the challenges of autonomous driving in complex road environments, the need for collaborative multi-tasking has been proposed. In the fields of natural language processing and recommendation algorithms, the use of multi-task learning networks has been proven to reduce time, computing power, and storage usage in multiple task coupling scenarios. Due to this characteristic of multi-task learning networks, in recent years, it has also been applied to visual-based road feature extraction. This paper proposes a decoder head structure combined with the FPN network and applies it to a YOLOv4-based multi-task learning road feature extraction network. Additionally, the paper optimizes the multi-task network algorithm through investigating the impact of multi-task weight settings. The effectiveness of the weight settings was also verified among similar algorithms. The experimental results obtained on the BDD-100K dataset show that the proposed structure has better accuracy while still ensuring real-time performance compared to similar methods. This paper's method provides new ideas and methodologies for vehicle autonomous road perception and high-precision map generation in visual-based autonomous driving processes.

A deep neural network model for road extraction of MLS LiDAR point cloud

LIU Jin, YANG Ronghao, WEN Wen, TAN Junxiang, LAN Qinglong, GAO Xiang, TANG Hong

2023, 0(12):  8-12,18.  doi:10.13474/j.cnki.11-2246.2023.0351

Asbtract ( )  

References | Related Articles | Metrics

PointNet++ has shown better performance than traditional methods in MLS LiDAR point cloud road extraction, but there are still the phenomena of over segmentation or under segmentation for road edge extraction.To address this issue, an improved neighborhood enhancement coding network E-PointNet++ is proposed. By introducing a neighborhood enhancement coding module before feature extraction, the connection between local neighborhood points is established to improve the network's road edge segmentation ability.Comparative experiments are conducted on two datasets, and E-PointNet++ shows significantly better performance than other methods, with accuracy, integrity and detection quality all exceeding 97%. This method performs robustly on different datasets and scenarios.

Front-of-vehicle road extraction method based on feature fusion difference of vehicle LiDAR

HE Guangming, HAN Shiyuan, CHEN Yuehui, ZHOU Jin, YANG Jun

2023, 0(12):  13-18.  doi:10.13474/j.cnki.11-2246.2023.0352

Asbtract ( )   HTML ( )   PDF (1514KB) ( )  

References | Related Articles | Metrics

In order to cope with the changing road environment during driving and divide the drivable area of the current road in front of the vehicle, this paper proposes a detection method for the road in front of the vehicle based on multi-feature fusion difference. This algorithm extracts the ground point cloud from the original point cloud by morphological filtering method, statistically summarizes the ground point cloud data to define the operation domain, divides the differential element size and starting point of different depths in the operation domain, fuses the characteristic parameters in the differential element, forms a feature matrix, solves the differential matrix, and performs threshold filtering, so as to realize the extraction of the point cloud in front of the vehicle. In this paper the extraction algorithm of the relevant road point cloud is compared to,which highlight its excellent performance and then the road extraction effect of different depths of the collected data is compared to prove the effectiveness of the algorithm.

Road network matching based on graph convolutional neural network

QI Jie, WANG Zhonghui, LI Yiyan

2023, 0(12):  19-24,44.  doi:10.13474/j.cnki.11-2246.2023.0353

Asbtract ( )  

References | Related Articles | Metrics

The current road network matching methods are highly subjective in determining the weighted relations and matching threshold and prone to matching errors. Therefore, a road network matching method combining graph convolutional neural network is proposed. First, four similarity measure factors of length, direction, distance and topology are selected as the feature factors of the road network matching model. Then the roads to be matched are transformed into a dual graph with roads as nodes and road connection relations as edges, thus to match road network through node classification. Finally, the road network matching is realized by building a graph convolutional neural network. The results show that the matching accuracy, recall rate, and F-value of this method are greatly improved compared with the traditional methods, and it can effectively solve the road network matching problem.

Enhanced lane line detection algorithm for curves based on Resa-CC

LU Weijia, LIU Zeshuai, PAN Yuheng, LI Guoyan, LI Huijie, CONG Jia

2023, 0(12):  25-30.  doi:10.13474/j.cnki.11-2246.2023.0354

Asbtract ( )   HTML ( )   PDF (1754KB) ( )  

References | Related Articles | Metrics

A curve enhanced lane detection algorithm based on cyclic feature fusion Resa-CC is proposed to address the issue of reduced accuracy in curve recognition caused by excessive curvature at road turns. This algorithm utilizes the shape priors of lane lines to capture the spatial relationships between rows and columns in image pixels, and fuse information to generate feature maps. The residual network is used as the main framework, and the encoder, decoder and attention mechanism modules are added. The Loss function adds curve structure constraints to improve the recognition accuracy of lane curves. The addition of the cyclic feature fusion module and the self attention mechanism module improved the accuracy by 3.41% and 1.1%, respectively, proving the effectiveness of the two modules. The accuracy of the Resa-CC algorithm can reach 96.83%, with an FPS of 35.68. The false detection rate FP and missed detection rate FN are 0.0315 and 0.0282, respectively. This indicates that the algorithm has high detection performance and can more accurately infer the position of the lane line in the curve when vehicles are driving.

SAR image change detection based on multi-scale fusion convolutional neural network

DUAN Yu, LIU Shanwei, WAN Jianhua, MUHAMMAD Yasir, ZHENG Shuang

2023, 0(12):  31-37.  doi:10.13474/j.cnki.11-2246.2023.0355

Asbtract ( )  

References | Related Articles | Metrics

The presence of speckle noise significantly affects the capability of synthetic aperture radar (SAR) image to recognize changing information. In order to enhance the accuracy of change detection, the influence of speckle noise must be adequately addressed. This study introduces a novel approach for change detection. Firstly, a context-aware saliency extraction method is employed to extract potential change regions and background information from the difference image. This process retains the main textural details of the image while removing background noise. A multi-scale channel attention module, the squeeze, expand, and excitation (SEE) module, is designed to address the issue of inadequate feature representation in current change detection methods. This module captures multi-scale information while emphasizing crucial details without introducing information redundancy.Building upon this foundation, a multi-scale fusion convolutional neural network called the squeeze, expand, and excitation network (SEENet) is proposed. SEENet connects three SEE modules through residual connections to achieve multi-level information utilization. Through experimentation on four real SAR datasets, the effectiveness of this method is validated.

Spatio-temporal evolution and monitoring assessment of carbon emissions in seven eastern provinces and cities based on nighttime light data

LIU Yaohui, LIU Wenyi, QIU Peiyuan, XING Huaqiao, LIU Yumin, WANG Qi, XING Xiaotian

2023, 0(12):  38-44.  doi:10.13474/j.cnki.11-2246.2023.0356

Asbtract ( )  

References | Related Articles | Metrics

As the main source region of carbon emissions in China, the analysis of the spatio-temporal evolution of carbon emissions in seven eastern provinces and cities is an important scientific basis for the formulation and implementation of carbon emission reduction strategies. This study aims to analyze the spatio-temporal evolution of carbon emissions in seven eastern provinces and cities from 2012 to 2021, and calculates the per capita carbon emission intensity and carbon emission intensity per unit of GDP. On this basis, a carbon emission and nighttime light fitting model based on “NPP-VIIRS-like” nighttime light data is constructed for carbon emission monitoring and evaluation in seven eastern provinces and cities. The results of the study show that:①Carbon emissions in seven eastern provinces and cities are generally on the rise, and the proportion of total carbon emissions in each province and city remains stable, with the fastest growth rate of carbon emissions in Jiangxi province.②The total carbon emissions show a pattern of “north>south”, with Shandong province and Jiangsu Province accounting for more than 50% of the total carbon emissions. ③The per capita carbon emission intensity shows an increasing trend in the southern region and a decreasing trend in the central region, while the northern region basically remains stable. The carbon emission intensity per unit of GDP shows a steady decreasing trend, especially in Shanghai and Zhejiang province, where the decrease is significant. ④The average correlation coefficient of the fitted model is 0.841 2, and the average relative error is decreasing, which indicate that carbon emission monitoring and evaluation can be effectively achieved based on nighttime lighting data. This study is of great significance to achieve industrial transformation and upgrading and sustainable development in seven eastern provinces and cities.

FCM-SBN-CVAPS multi-scale target change detection based on multi-spectral heterologous remote sensing image

WU Jinsha, YANG Shuwen, LI Yikun, ZHAO Zhiwei, ZHENG Yao, FU Yukai

2023, 0(12):  45-50.  doi:10.13474/j.cnki.11-2246.2023.0357

Asbtract ( )  

References | Related Articles | Metrics

In remote sensing image change detection, FCM-SBN-CVAPS method can effectively deal with the problem of mixed pixels. However, due to the large differences between heterogeneous images, it has limitations for heterogeneous image change detection. In order to improve the accuracy of change detection, in fuzzy C-means (FCM), simple Bayesian network, (SBN) and change vector analysis in posterior probability space (CVAPS), In this paper, a multi-scale change detection method of FCM-SBN-CVAPS for heterologous images is proposed. Firstly, image quality is improved by image enhancement. Secondly, in order to improve the accuracy of the posterior probability vector calculation of FCM-SBN, and effectively judge the areas with different spectra of the same object and the same spectrum of foreign objects, the subclasses of ground object samples are combined to form compound type ground object samples, and the large target change detection of FCM-SBN-CVAPS is realized. At the same time, subclass samples are used to redetect the missed areas to realize small target change detection, and the change information of different scales is superimposed to obtain the final change detection result. Finally,two groups of heterosource image data are used to compare and verify the proposed method. The results show that the proposed method can reduce the false detection rate and missed detection rate, and the overall accuracy and Kappa coefficient are higher than the comparison method.

Evaluation of geological hazard vulnerability in Xining city based on InSAR and informativeness-hierarchical analysis coupled modeling

HU Xiangxiang, MING Lulu, WU Tao, LIU Baokang, PANG Dongdong, YIN Jixin, SONG Bao, KE Fuyang

2023, 0(12):  51-56,75.  doi:10.13474/j.cnki.11-2246.2023.0358

Asbtract ( )  

References | Related Articles | Metrics

To accurately and efficiently evaluate the landslide susceptibility area within the territory, it provides a reference for landslide disaster warning and prevention in Xining city. In this paper, SBAS-InSAR technology is used to obtain the surface deformation information of Xining city during the five years of 2018—2022. The deformation information obtained is divided into magnitudes as an evaluation factor of the susceptibility evaluation model, which is combined with seven evaluation factors, such as elevation, slope, and stratigraphic lithology, to quantitatively evaluate the susceptibility of landslides in the study area based on the informativeness-hierarchical analysis method. Finally, the evaluation results were validated using the existing geohazard point data sets. The results show that the evaluation results of geohazard susceptibility obtained in this paper are more consistent with the distribution of existing geohazard sites, and most of the areas where the recorded geohazard sites are gathered are located in the medium-high susceptibility zone of the location delineated in this paper. According to the ROC curve accuracy results, the coupled model AUC value is 0.91.It is feasible to utilize the method proposed in this paper to carry out disaster susceptibility evaluation, and the results of the study can provide a basis for the comprehensive prevention and control measures of landslides in Xining city.

Cropland recognition in southern hilly areas under a hybrid U-Net model

WU Ruijiao

2023, 0(12):  57-62,111.  doi:10.13474/j.cnki.11-2246.2023.0359

Asbtract ( )   HTML ( )   PDF (11831KB) ( )  

References | Related Articles | Metrics

In order to solve the difficulties and low accuracy of sloping cropland identification caused by scattered, fragmented and irregular cropland in southern hilly areas, based on the U-Net, the efficient channel attention(ECA) and the attention gate (AG) dual attention mechanism are introduced, and a hybrid U-Net model is proposed. This model is applied to extract cropland from WorldView-2 satellite images in Nan'an of Fujian province in 2021. Experiment shows that the hybrid U-shaped network model has achieved a good accuracy of 93.42%, which is better than a single-attention mechanism model (ECA U-Net and U-Net), and the accuracy has increased by 9.75% and 19% respectively. The average F1 scores of cropland in the hybrid U-Net model are 0.921 2, 0.902 5 and 0.932 2 respectively in the mountainous, semi hilly and plain test areas, especially in the mountainous and semi hilly areas. On this basis, the spatial distribution of cropland in Nan'an is analyzed, which provided effective technical support for abandoned hillside fields for grain cultivation, rational adjustment of slope cropland exceeding 25 degrees, and effective control of cropland quantity.

Multi-type pavement distress segmentation methods in complex scenarios

ZHANG Zaiyan, SONG Weidong, CHEN Zhaoxue

2023, 0(12):  63-69.  doi:10.13474/j.cnki.11-2246.2023.0360

Asbtract ( )  

References | Related Articles | Metrics

Automatic segmentation of various types of pavement distresss is of great significance for improving the level of highway maintenance management. The lack of large-scale, diverse-scenario, and multi-type training data for pavement distresss has reduced the generalization ability of deep learning models in complex scenarios, and limited the development of engineering applications for pavement distress extraction algorithms. To address this issue, this paper collects and establishes a dataset called CPCD for the segmentation of multiple types of pavement distresss, aiming to accurately segment pavement distresss from CCD images. The new dataset covers 7 common types of pavement distresss, including negative samples with texture-similar noise, totaling 6967 images. Based on this dataset, a distress segmentation convolutional neural network called CBAM-HRNet, which combines attention mechanisms with high-resolution networks is proposed to improve the accuracy of fine-grained distress segmentation. Experimental results show that the F1 score and mIoU of the proposed model on the CPCD dataset are 91.30% and 84.64% respectively. Its performance is significantly better than other mainstream convolutional neural network models with sequential structures. The research findings will provide certain support for the research on highway automation detection and sustainable development in our country.

Intelligent identification and location of defects in water supply pipeline based on improved YOLOX algorithm

SU Changwang, HU Shaowei, ZHANG Haifeng, PAN Fuqu, SHAN Changxi

2023, 0(12):  70-75.  doi:10.13474/j.cnki.11-2246.2023.0361

Asbtract ( )   HTML ( )   PDF (4266KB) ( )  

References | Related Articles | Metrics

To solve the problem of difficult and slow real-time automated detection of defects in water supply pipelines, a new intelligent identification and positioning method for water supply pipelines is proposed based on a dataset of pipeline defect data collected from actual engineering projects. The new YOLOX algorithm model, which incorporates an attention module, is developed and used for algorithm training and prediction using a dataset of video frames. Test results show that the YOLOX algorithm model with attention mechanism achieved an average testing accuracy of 94%, a mAP value of 84%, and an average recognition speed of 16 m/s. Additionally, compared with three other commonly used algorithm models (YOLO V3 and Fast R-CNN), the new model showed the best overall performance. This proposed model can also be applied to real-time video detection, providing an efficient and accurate detection technology and method for the intelligent identification and positioning of defects in water supply pipelines.

Analysis of groundwater potential in Jiangyou city by using spatial hierarchy overlay method

ZHANG Zhenping, LI Yingbing, ZHANG Pei

2023, 0(12):  76-80,177.  doi:10.13474/j.cnki.11-2246.2023.0362

Asbtract ( )  

References | Related Articles | Metrics

Groundwater potential analysis is used to identify and protect groundwater resources and reduce the cost of water source exploitation. The traditional groundwater potential analysis method is mature in technology and has a simple algorithm, but it has the problems of a long mining period and low prediction accuracy. This paper proposes a spatial hierarchy overlay method (SHOM) for groundwater potential. Firstly, it uses spatial analysis methods to extract hydrogeological data, and selects multiple evaluation factors based on the mechanism of groundwater concentration. Then it uses AHP to determine the weight of each factor to replace the traditional average weighting or subjective weighting process. Finally, a geographical unit is established for spatial overlay analysis to complete groundwater potential evaluation. In this paper, Jiangyou city, Sichuan province, serves as the study area. Based on the proposed SHOM analysis method, a groundwater potential model is constructed according to 12 hydrogeological characteristics, such as elevation, slope, distance from the river, TWI, TRI, etc., and regional groundwater potential is evaluated. The results show that the areas with low, medium, and high groundwater potential in Jiangyou city account for 12.97%, 61.38%, and 25.65%, respectively, and have been verified with 72 groundwater wells.

Regionally geographically weighted regression method

WANG Zengzheng, ZHANG Fuhao, ZHAO Yangyang, QIU Agen

2023, 0(12):  81-87.  doi:10.13474/j.cnki.11-2246.2023.0363

Asbtract ( )  

References | Related Articles | Metrics

Geographically weighted regression (GWR) serves as a potent approach to discern spatially continuous heterogeneity, premised on the assumption of proximity correlation. However, in practical applications, particularly within the socio-economic domain, the presence of “near-heterogeneous” spatial discrete is frequently observed. Consequently, the challenge of concurrently detecting spatially discrete and continuous heterogeneity to enhance the estimative precision of GWR warrants further investigation. In this study, we introduce a regionally geographic weighted regression (RGWR) analysis methodology, which effectively filters observation points by devising a regional spatial weight computation strategy, refining the spatial kernel function, optimizing spatial weights, and mitigating the impact of “near-heterogeneous” observation points. We utilize housing sales prices in Wuhan as the empirical case, examining the data from three perspectives: regional factor effectiveness, model performance, and model fit. The findings reveal that incorporating regional impact factors considerably enhances model accuracy under both fixed and adaptive bandwidths. Specifically, the regional impact factors stemming from educational determinants yield the most substantial improvement in model accuracy. Simultaneously, under fixed bandwidth conditions, the model's R² value increases by 21.84%, while the MSE rises by 37.09%. This evidence underscores the model's heightened accuracy upon considering regional influence factors, thereby substantiating the effectiveness of the proposed method.

Combining recursive feature elimination and 2D CNN for landslide susceptibility evaluation

ZHANG Pei, LI Yingbing, ZHANG Zhenping, HU Lutai

2023, 0(12):  88-93.  doi:10.13474/j.cnki.11-2246.2023.0364

Asbtract ( )  

References | Related Articles | Metrics

In response to the problem that the traditional landslide susceptibility analysis methods only consider the impact factor information of the landslide point itself and ignore the surrounding spatial information, a method that combines recursive feature elimination and a 2D convolutional neural network is proposed. Firstly, the recursive feature elimination method is used to rank and filter the landslide impact factors. Subsequently, the 2D feature factor set is cropped and fed into a 2D CNN with L2 regularization, Dropout, and other optimization methods, and the spatial information around the landslide is taken into account to predict the landslide susceptibility while ensuring the prediction accuracy and generalization ability of the model. In this paper, the Jiuzhaigou area is taken as the experimental area, and 14 relevant factors such as elevation and lithology are selected as landslide-influencing factors to predict the probability of landslide occurrence and draw a landslide susceptibility map. Finally, a logistic model and three SVM models with different kernel functions (linear kernel function, radial basis kernel function, and sigmoid kernel function) are used for comparison and validation. The experimental results show that the proposed method has the highest accuracy and AUC, which proves the validity and reliability of the proposed method.

The construction of the deep-sea time variant thermohaline model in the North Pacific Ocean by combining CTD, seabed terrain and ARGO data

ZHANG Jinhui, LI Shanshan, YANG Guang, FAN Diao, LING Qing

2023, 0(12):  94-101,126.  doi:10.13474/j.cnki.11-2246.2023.0365

Asbtract ( )  

References | Related Articles | Metrics

Faced with the fact that the measured data of deep-sea temperature are insufficient, this paper redivided the experimental sea area of the North Pacific ocean according to the characteristics of CTD temperature profile changing with sea depth. Combined with the seabed terrain and Argo data, the deep-sea monthly grid temperature model of the North Pacific ocean from 2005 to 2020 is constructed,and the deep ocean steric sea level change is inversed. The experiments results show that: ①Compared with other mathematical model, the difference between the mean mathematical model of deep-sea temperature profile constructed in this paper and the measured data of CTD is 1～2 orders of magnitude smaller, which can more accurately reflect the characteristics of deep-sea temperature profiles changing with sea depth.in various regions. ②The maximum difference between temperature model data and CTD measured data, EN4 do not exceed 0.20℃ and 0.60℃, the average do not exceed 0.03℃ and 0.50℃, and the standard deviation not exceed 0.06℃ and 0.002℃.③ Based on the temperature model and EN4, the deep ocean steric sea level change in the North Pacific ocean is basically consistent, in which the rising trend from 2005 to 2010 is 0.52±0.09 and 0.73±0.11 mm/a, and the rising trend from 2010 to 2020 is 0.02±0.03 and -0.01±0.01 mm/a,which is consistent with the research conclusion based on heat content change in relevant literature.The rising trend in the whole study period is 0.11±0.17 and 0.09±0.11 mm/a. This shows that the temperature model data constructed in this paper has a certain reliability, which have a certain reference value for refining the genetic changes of regional sea level balance equation.

Forest canopy height and biomass estimation based on LiDAR satellite (GEDI) in Guangdong province

WU Zhenjiang, ZHANG Jiahua

2023, 0(12):  102-105.  doi:10.13474/j.cnki.11-2246.2023.0366

Asbtract ( )  

References | Related Articles | Metrics

Forest canopy height and biomass estimation play an important role in estimating forest carbon expenditure. In this study, the forest canopy height and biomass in Guangdong province use the global ecosystem dynamics survey (GEDI) LiDAR satellite as the data source, regression tree and Kerry kin interpolation algorithm, respectively. The results show that the height of trees in Guangdong province is generally between 10 and 20 m, accounting for more than 50%. The tree height high value occurs in Shaoguan, Zhaoqing and other cities in northern Guangdong province, and the tree height is generally 15~20 m, while the average tree height in Zhanjiang city is the lowest, generally less than 10 m. The maximum forest biomass in Guangdong province is 335.85 t/hm², the minimum value is 5.25 t/hm², and the average value is 98.27 t/hm².The areas with high value of forest biomass are mainly distributed in the eastern and western Guangdong province, while the forest biomass is lower in the plain and urbanized areas of Guangdong province. The results provide a scientific basis for estimating carbon absorption of forest ecosystem in Guangdong province.

Application of InSAR deformation monitoring in mining subsidence damage identification and stability evaluation

CHEN Ranli, WANG Hongjun, DIAO Xinpeng, ZHANG Baojin

2023, 0(12):  106-111.  doi:10.13474/j.cnki.11-2246.2023.0367

Asbtract ( )  

References | Related Articles | Metrics

This paper illustrates the application of InSAR technology in mining subsidence by listing three engineering cases. Firstly, the InSAR technology is compared with the third-class leveling to clarify the accuracy and reliability of InSAR technology in the field of mining deformation monitoring. Secondly, SBAS-InSAR technology is used in a closed mine in Huainan to obtain the time series deformation information, and the solution results are compared with the goaf stability evaluation index, concluding that the surface of the mining area is stable.Finally, based on the research background of some houses and buildings in a village with different degrees of damage, and on the basis of conventional methods to obtain the scope of mining impact, the SBAS technology and 8 Sentinel-1A data are further used to confirm that the village surface has been in a stable state. The results show that InSAR deformation monitoring technology can provide a powerful scientific basis for mining subsidence damage tracing and settlement stability judgment.

Monitoring of vegetation covered slopes based on 3D laser

WANG Dejun, WAN Tianbao, SUN Xiaodong, XU Jingli, DU Zitao

2023, 0(12):  112-115.  doi:10.13474/j.cnki.11-2246.2023.0368

Asbtract ( )  

References | Related Articles | Metrics

It is difficult to obtain high-precision ground point cloud data for vegetation covered slopes, which leads that it's difficult to accurately monitor the surface deformation of slopes. In this paper, a method is proposed to extract the deformation of slope by using the point cloud data on the sensor surface distributed on the slope body. The shape of GNSS antenna radome established on the slope is a spherical equipment with gradually changing radius. The circle center coordinates and radius are fitted by the point cloud slice, and the accuracy of the point cloud is evaluated by the fitting residual and the posterior mean square error, the quality control is realized. Through the comparison of radius in two phases, the matching of points with the same name is realized, and the purpose of point monitoring is achieved according to the coordinates of the circle center in two phases. Surface monitoring can be realized according to the steps on the slope and the peripheral equipment of the rain gauge. The measured data shows that the results of 3D laser scanning is consistent with the ones of GNSS monitoring, and the method can extract millimeter level deformation.

Construction of city level 3D realistic geospatial scene using tilt photography and laser scanning technology

LUO Zhenwei, LI Xiao, LIU Chengcheng, ZHANG Yong

2023, 0(12):  116-120.  doi:10.13474/j.cnki.11-2246.2023.0369

Asbtract ( )  

References | Related Articles | Metrics

The advancement of the “Real 3D China” construction plan is increasing demand for large-scale, high-definition city-level real 3D production. Starting with the top-level design and combining it with oblique photography, laser scanning, and other technologies, this research builds a technical framework for the entire life cycle of urban-level real scene 3D construction, using Chengdu's Tianfu New district as the research area to validate the technical framework's feasibility. The research results indicate that the final 3D real-scene model's accuracy meets the design requirements, the expression quality meets the requirements, and the logic is consistent. Standardizing the technical process has reduced production costs significantly, and it also served as a reference for real-world 3D construction research in other cities.

Research on fine modeling and visualization of large-scale central city

HUA Anzhong, SUN Hao

2023, 0(12):  121-126.  doi:10.13474/j.cnki.11-2246.2023.0370

Asbtract ( )  

References | Related Articles | Metrics

The real 3D model of UAV will be distorted and deformed when the data is missing, the precision of the model is not high,and it can not be directly used in the management application of smart cities due to the modeling mechanism and other reasons. This paper takes the fine modeling of the 100 square kilometers central urban area of Jiangning district, Nanjing as the research object, use multi-source heterogeneous data to build a 3D model of large scenes,completed the fine modeling and 3D visualization of various infrastructure based on the 3D model. The results show that the constructed refined model has high mathematical accuracy, exquisite model expression and excellent integrity,the 3D visualization system can carry the rapid loading and operation of massive 3D model data, and can provide services and applications for the management of smart cities,The research provides certain reference value for the construction of large-scale 3D refined city models.

Carbon sink change in Datong coalfield based on hyperspectral image of Zhuhai No.1

YUAN Yuan

2023, 0(12):  127-131.  doi:10.13474/j.cnki.11-2246.2023.0371

Asbtract ( )  

References | Related Articles | Metrics

Due to the mining history of Datong coal field for many years, it has destroyed the agriculture, forestry and grassland, and become a fragile ecological environment area, resulting in the change of carbon cycle in the coal field area. With the green development in recent years, the ecological environment has improved, and the carbon sink has increased. Taking Datong coal field as the research area, based on the hyperspectral remote sensing image data of Zhuhai No.1 in 2020 and 2021, this paper studies the carbon sink change in the coal field area in the two years from the perspective of land use change. The conclusions are as follows: ①the land use structure of Datong coal field area has not changed significantly in the two years, and the grassland, forest land and cultivated land have increased by 2.57 km², 0.71 km² and 0.11 km² respectively; ②the carbon sink of Datong coal field in 2021 has increased by 30 000 t CO₂ compared with that in 2020. The ecological environment of the coal field has gradually improved.

Application of nap-of-the-object photogrammetry in the 3D modeling of tangible cultural heritage

LUO Xiaodan, LAI Mingzhi, LU Yan, Lü Xinqiang, LING Congcong, HUANG Yu

2023, 0(12):  132-135,152.  doi:10.13474/j.cnki.11-2246.2023.0372

Asbtract ( )  

References | Related Articles | Metrics

As a case study, Huashan rock paintings in Ningming county, Chongzuo city, Guangxi are analyzed by using UAV close photogrammetry technology to research 3D model modeling of tangible cultural heritage. The goal is to explore digital archival technology and methods suitable for 3D modeling of cultural heritage, including rock paintings, carvings, and ancient buildings. The study reveals that the initial terrain model is created through the collection of initial terrain image data using close-range photogrammetry technology on unmanned aerial vehicles (UAVs), the planning of refined routes based on the initial terrain model, and the implementation of high-resolution photogrammetry. The resulting refined 3D model has high resolution, clear texture, and accurately restored character information. The research findings can offer precise guidelines on 3D modeling technology for related fields that involve acquiring geospatial information, like safeguarding natural and cultural sites. This technology holds promising potential for wider circulation and usage.

Classification of coastal wetlands in the Pearl River Estuary using Zhuhai-1 hyperspectral imagery and XGBoost algorithm

LIU Yanjun, LIU Kai, CAO Jingjing

2023, 0(12):  136-141.  doi:10.13474/j.cnki.11-2246.2023.0373

Asbtract ( )  

References | Related Articles | Metrics

Remote sensing classification of wetlands is still challenging due to the diversity of wetland types and complex composition. Taking the Pearl River Estuary as the study area, based on Zhuhai-1 hyperspectral imagery, we extracted the wetland type information with the spectral features, shape features, texture features, and spectral indices, using the eXtreme gradient boosting (XGBoost) algorithm, and compared with support vector machine (SVM) and random forest (RF). Results showed that Zhuhai-1 imagery can be used to identify wetland types accurately. Among three machine learning algorithms, the XGBoost gave the best wetland classification effect (OA=87.2%, Kappa coefficient=0.84). Moreover, the selected features gave higher classification accuracy, which verified the importance of feature selection for Zhuhai-1 imagery. This study proposed a new method suitable for large-area wetland classification, which can provide a practical technical reference for wetland resource investigation, protection, and development.

Application of UAV oblique photogrammetry technology in buildable photovoltaic resource survey

LIU Li, WEI Haixia, ZU Weiguo, TAN Jinshi

2023, 0(12):  142-146.  doi:10.13474/j.cnki.11-2246.2023.0374

Asbtract ( )  

References | Related Articles | Metrics

Aiming at the current problems of high work intensity, low mapping efficiency, and difficult to measure elements, this paper proposes an auxiliary photovoltaic roof area measurement method based on UAV tilt photogrammetry. This paper studies on oblique photography, three-dimensional modeling, mapping and evaluates the accuracy of the measurement results. The results show that the RSME of three-dimensional model in the plane position is 3.9 cm, the error in the elevation is 4.3 cm, the error in the corner of the building is 4.3 cm, the error in the side length of the building is 3.1 cm, and the area of the building is smaller than the second-level tolerance accuracy of the urban commercial building, all the accuracy detected in the case meets the requirements of national standards, so it is feasible to use this method to measure the area of photovoltaic roofs.

Applications of new surveying and mapping technology in urban component investigation

ZHOU Changjiang, ZHANG Chenhui

2023, 0(12):  147-152.  doi:10.13474/j.cnki.11-2246.2023.0375

Asbtract ( )  

References | Related Articles | Metrics

In order to address the issues of high production cost, high difficulty and low efficiency in traditional urban component investigation, multiple new surveying and mapping technologies with mobile measurement technology as the core have been applied. The technologies adopt a field collection method that combines vehicle-borne mobile mapping system, portable mobile scanning system and mobile terminal software, enabling rapid acquisition of component elements with the help of efficient component element extraction software. Furthermore, the integration of census results into the digital twin platform has laid the foundation for further application of component data.

Secret target automatic recognition and decryption method for real scene 3D model texture

XU Haiyan, GUO Weiren, LI Demin, HAO Jun, XU Gang

2023, 0(12):  153-158.  doi:10.13474/j.cnki.11-2246.2023.0376

Asbtract ( )  

References | Related Articles | Metrics

Real-scene 3D is an important part of the country's new infrastructure and has a wide range of applications in various industries. How to promote the maximum sharing of 3D data under the premise of safety has become the demand of real-scene 3D applications. Aiming at the problem of secret-related sensitive targets in real-scene 3D textures, the traditional decryption process that relies on manual retrieval of sensitive targets and images processed by editing tools is not efficient. This paper proposes an automatic recognition and decryption method for real scene 3D model texture combined with deep learning. Firstly, search the 3D model and texture image containing the secret target through the secret areas POI; then automatically identify the sensitive target in the texture image based on the YOLOv5s network model, and use GrabCut to effectively extract the target; finally, based on the multi-scale Patch match for the texture image block make repairs. It shows that the target recognition accuracy of the method is 95.3%, which is more than 40% compared with manual processing when the whole process is used. It effectively extracts sensitive targets to achieve fast decryption, and promotes the safe sharing of real-scene 3D model data.

Extracting rail profile and measuring wear based on machine vision

ZHAO Jing

2023, 0(12):  159-163,177.  doi:10.13474/j.cnki.11-2246.2023.0377

Asbtract ( )  

References | Related Articles | Metrics

With the rapid development of China's high-speed railways, traditional contact-based rail wear detection methods have limitations in measurement efficiency and are easily affected by rail deformations and surface damages, failing to meet the demands of the rail maintenance. To quickly and accurately measure the rail profile and assess rail wear, a contour extraction method based on machine vision is proposed. A monocular machine vision system for rail profile extraction has been established. After positioning with a free planar target, a two-level rail profile extraction method based on structured light color and the Steger algorithm is employed. A double circle fitting rail profile matching algorithm based on particle swarm optimization is introduced to align the measured profile with the standard profile and derive the wear value. Experimental results demonstrate that this system achieves a detection accuracy of 0.128 mm and offers high measurement efficiency.

Optimization of the BDS-2 triple-frequency un-combined PPP random model without inter-frequency clock deviation correction

ZHOU Changjiang, YU Haifeng, WANG Linwei, LEI Yunping, YUE Caiya

2023, 0(12):  164-168.  doi:10.13474/j.cnki.11-2246.2023.0378

Asbtract ( )  

References | Related Articles | Metrics

In view of the inter-frequency clock deviation (IFCB) of BeiDou-2 Navigation Satellite System (BDS-2) will lead to systematic deviation in multi-frequency precision point positioning (PPP), an optimization scheme for a triple-frequency un-combined PPP stochastic model suitable for BDS-2 without IFCB correction is presented. The method of optimal stochastic model building is objectively evaluated and analyzed through data from 30 monitoring stations distributed in the Asia-Pacific region. The results show that the BDS-2 third frequency carrier phase observed variance should be amplified to 2.0 to 4.0 times of the original variance. In order to further verify the usability of this stochastic model, the static and dynamic solutions are implemented for six additional stations. Both the positioning accuracy and convergence time shown the same effect as the real IFCB correction, and it is better than the BDS-2 triple-frequency un-combined PPP which does not take into account the influence of IFCB.

Evaluation and visualization of soil heavy metal pollution

YU Qianhui, WANG Honglin, SHI Jiangchen, FU He, ZHANG Lu, ZHAI Xinya

2023, 0(12):  169-173.  doi:10.13474/j.cnki.11-2246.2023.0379

Asbtract ( )  

References | Related Articles | Metrics

For the evaluation of heavy metal pollution in soil, researchers often use only one assessment model, and can not choose the best one; In the case of limited access to soil data, data availability and adjustability are poor; Direct evaluation of heavy metal pollution based on a small number of sampling point data is not conducive to the evaluation of the overall impact effect and other issues. In this paper, the evaluation and visualization of heavy metal pollution in soil are studied. The data of sampling points are transformed by Kriging method. Combined with single factor evaluation model, Nemerow index model and weighted average index model, the formula is programmed to achieve rapid evaluation of heavy metal pollution. Combined with open-source WebGL technology, the highly efficient visualization is conducted in the form of pollution degree classification. The research results show that the method proposed in this paper not only could enable users to select evaluation models and interpolation accuracy as required, but also have high data availability, which is conducive to evaluating the overall pollution situation in the study area and obtaining the optimal evaluation results of the area.

Digital protection of ancient tombs based on 3D laser scanning technology

LIU Xiuhan, ZHANG Pengdong

2023, 0(12):  174-177.  doi:10.13474/j.cnki.11-2246.2023.0380

Asbtract ( )  

References | Related Articles | Metrics

Ancient tombs are precious heritage of the development of human civilization, and they are usually with extremely high historical, cultural and artistic values. Affected by various natural and man-made factors, ancient tombs are often destroyed to some extent, therefore, the digital protection of ancient tombs are urgently needed. It is usually difficult to collect the underground data by traditional tools such as total station and RTK given that ancient tombs are generally hidden underground. To this end, this paper uses the Trimble X12 3D laser scanner to realize the accurate collection of point cloud data of ancient tombs. Then, the internal and external real scenes of the ancient tombs are truly restored through 3D modeling, and on this basis, the digital protection application of the ancient tombs is studied. The research results show that this technology can provide a solid data foundation for the restoration and protection of ancient tombs.

Terrain level geographic scene construction based on Leica CityMapper mixed data

ZHU Lei, ZHAO Fei, WANG HongChang

2023, 0(12):  178-180.  doi:10.13474/j.cnki.11-2246.2023.0381

Asbtract ( )  

References | Related Articles | Metrics

This article proposes a terrain level geographic scene construction method based on Leica CityMapper mixed data, elaborates on the main steps and key technologies involved in this method, and verifies and analyzes it through a provincial-level real-life 3D construction example. The results show that this technology has the advantages of high elevation accuracy, consistency between texture and model time and saes updating cost. This technology has obvious advantages in the updating of terrain level geographical scenes, especially in the updating of banded areas (new high-speed railways, highways, canals and so on),which has high application value.