A Case Study on the Application of 3D Scanning Technology in Deformation Monitoring of Slope Stabilization Structure

Abstract

Traditional deformation monitoring suffers from issues such as the point-based representation of surfaces and low measurement efficiency. Moreover, the majority of researchers study the deformation of slopes using methods such as 3S technology, synthetic aperture radar interferometry, distributed fiber optic sensing technology, etc. Based on this, a slope stabilization structure deformation monitoring method based on 3D laser scanning technology is proposed. First, with the slope stabilization structure of Caihong Road as the engineering background, point cloud data of the slope stabilization structure is obtained using a Trimble SX10 device. Second, the point deformation, overall deformation, and line deformation of the two-phase slope stabilization structure point cloud data are analyzed. Finally, the measurement accuracy of the 3D laser scanning technology is evaluated. The results show that the deformation analysis of points, lines, and surfaces can complement each other, thereby comprehensively assessing the situation of slope stabilization structure deformation. Moreover, the maximum displacement value in the deformation of points, lines, and surfaces is 8.52 mm, which does not exceed the standard, and 93.61% of the point deformation is between −0.76~0.92 mm, indicating that the slope stabilization structure is in a safe and stable state. The independent sample t-test has a test statistic of t = 2.074, verifying that the 3D laser scanning technology and the total station measurement accuracy are highly consistent and can meet the needs of actual engineering. The results of this study can provide a reasonable theoretical and methodological reference for analyzing similar engineering deformation monitoring in the future.

1. Introduction

During their operational period, slopes can undergo varying degrees of deformation. When the deformation exceeds the defined limits, it can trigger disasters such as landslides, mudflows, and collapses, events that frequently occur and are widely distributed within the field of civil engineering. To ensure slope stability, it is necessary to monitor the deformation status of slopes [1,2]. The concept of deformation monitoring was proposed in the 1960s, and the dynamic monitoring of slopes can not only prevent landslides but also predict and analyze deformation trends, thus ensuring the safety of people’s lives and property and safeguarding the security of national economic construction [3].

The first commercial Zeiss RecElta14 theodolite in 1968 and its later updated and developed traditional monitoring instruments suffered from limitations such as single-point analysis, poor reliability, low work efficiency, and high safety risks [4,5,6]. Currently, 3D laser scanning technology is the most promising deformation monitoring technique [7]. Utilizing this technology allows for the local or overall monitoring of target areas, thereby obtaining complete panoramic point cloud data and a real-time grasp of the dynamic changes in the slope, avoiding the limitations of single-point monitoring and offering a high scanning efficiency and non-contact measurement advantages [8]. Researchers domestically and abroad have monitored the deformation of slopes and landslides through different methods such as “3S” technology [9,10], synthetic aperture radar interferometry [11,12,13,14,15], distributed optical fiber sensing technology [16,17,18,19], and photogrammetry [20,21,22]. Gili et al. [9] monitored landslides in the eastern Pyrenees of Spain using GPS technology, with results demonstrating good precision. Additionally, Li et al. [10] established a 3D landslide monitoring system based on GIS concepts, providing an analytical platform for landslide monitoring results. Gischig et al. [11] conducted five measurements of the rock slope stability in Randa, Switzerland, through ground differential interferometric synthetic aperture radar (DInSAR), with results aligning well with the measurements from total station instruments. Barla et al. [12] used GBInSAR monitoring technology to observe deformation in the gravity slopes of the Aosta Valley in northwestern Italy, obtaining multi-temporal surface deformation on landslides and validating radar monitoring results through comparison with automatic total station topographic measurements. Li et al. [13] combined deformation data with GIS organically through GB-InSAR, displaying intuitively and in real time deformation information, providing strong decision support for coal mine slope deformation monitoring. Su et al. [14] proposed a near-real-time GB-InSAR method for monitoring the surface deformation of slopes, and validated the method using measured data, demonstrating excellent results. Tan et al. [15] proposed a landslide time prediction method based on the time series monitoring data of micro-deformation monitoring radar, and the results showed that the method can identify imminent landslides and landslide characteristics in advance. Sun et al. [16] used distributed optical fiber sensing technology for the distributed collection, characterization, and stability analysis of multi-field information on slopes, even proposing corresponding slope evolution laws and verifying DFOS monitoring results with the landslide case in Majiagou. Zhu et al. [17] studied the feasibility and economic analysis of using a new type of distributed optical fiber transmitter for landslide monitoring, discovering a high monitoring efficiency and stability in slope and tunnel construction civil engineering monitoring. Acharya et al. [18] emphasized the important considerations for deformation sensing when installing fiber optic cables in boreholes/shallow trenches and the coupling mechanism between ground anchors and cables. Embedding fiber optic cables in shallow trenches or burying them in boreholes can detect early signs of failure before a collapse. Zhang et al. [19] demonstrated that an inclinometer based on fiber optic Bragg grating has a high monitoring accuracy, and the data can be transmitted without being affected by electromagnetic interference, enabling three-dimensional deformation monitoring in space. Zhao et al. [20] developed a new non-contact monitoring method based on a photogrammetry system for measuring slope surface displacement, monitoring slope surface displacement through cameras and measuring plates, and they proposed a new marker design method to improve measurement precision. Li et al. [21] monitored mountain landslides using close-range photogrammetry technology and used simulation data to verify that the method can monitor quickly and accurately. Li et al. [22] proposed a high-precision, low-cost, and automated method for monitoring slope deformation, which can be used in conjunction with drones and industrial photogrammetry techniques.

In summary, the current research on slope deformation monitoring mainly utilizes methods such as 3S technology, synthetic aperture radar interferometry, and distributed fiber optic sensing technology to study slope deformation, while there has been relatively little research on slope stabilization structure deformation analysis using 3D laser scanning technology. This paper’s novelty lies in two aspects: on the one hand, it proposes analyzing slope deformation from the three perspectives of single point deformation, overall deformation, and line deformation to avoid errors caused by single point measurement; on the other hand, it proposes comparing and analyzing data obtained by 3D laser scanning technology and total station to verify the reliability of 3D laser scanning technology in slope deformation monitoring. Therefore, it is of great significance to use 3D laser scanning technology for slope deformation monitoring research.

2. Overview of Point Cloud Registration Algorithms

2.1. Extracting FPFH Point Cloud Features

A Trimble SX10 device was used in this study to collect point cloud information from the slope stabilization structure, and the key point information in the slope stabilization structure point cloud was extracted. The geometric features of points and local areas were described by introducing a point feature histogram (PFH) and fast point feature histogram (FPFH). The information provided by FPFH has translation and rotation invariance, which can have robustness and repeatability in detecting sampling density and noise points while still retaining most of the discriminative power of PFH [23].

To begin, the correlation between a query point ($P_{\mathrm{q}}$) and its neighboring points in the point cloud of the slope stabilization structure is first computed. The angular deviation ($\alpha ,\Phi ,\theta$) between a paired source point ($P_i$) and a target point ($P_j$) is expressed using a simplified point feature histogram (SPFH), and the calculated results are stored in the feature histogram of each point. The computation process is described by Equation (1).

$$\left\{\begin{array}{c}\alpha =v\cdot n_j\\ \Phi =(u\cdot (P_j-P_i))/d\\ \theta =\arctan (w\cdot n_j,u\cdot n_j)\end{array}\right.$$

(1)

where $n_i$ and $n_j$ are the normal vectors of the source point $i$ and target point $j$; $d$ represents the Euclidean distance between the two points; and $u,v,w$ denote the local coordinate system ($u=n_i,v=(P_j-P_i)\times u,w=u\times v$) defined at the source point $i$.

Additionally, for a given point cloud P with n points, the theoretical computational complexity of calculating the feature histogram is denoted as $O(n\cdot k)$. The influence of the local coordinates of two points and the range of FPFH are illustrated in Figure 1.

Finally, the neighborhood of the query point $P_{\mathrm{q}}$ is recalculated. The $SPFH$ is computed using the feature elements between the neighboring point $P_k(k=1,2,3,4,5\cdots )$. The resulting $FPFH$ of point $P_q$ is then used as the basis for point cloud registration, which is obtained through Equation (2).

$$FPFH(P_q)=SPFH(P_q)+\frac{1}{k}{\displaystyle \sum _{i=1}^k\frac{1}{{\omega }_k}}\cdot SPFH(P_k)$$

(2)

where $FPFH(P_q)$ represents the $FPFH$ of the query point; $SPFH(P_q)$ represents the $SPFH$ of the query point; $k$ denotes the number of neighboring points in P; and ${\omega }_k$ represents the weight.

2.2. Sample Consensus Initial Alignment (SAC-IA)

After extracting the feature information from the slope stabilization structure point cloud, the SAC-IA algorithm is utilized for optimization and computation by selecting slope stabilization structure point clouds with similar attribute relationships. The coarse registration process using SAC-IA is as follows:

(1)

Select sampling points in the source point cloud, ensuring that the Euclidean distance between sampling points is greater than a predefined threshold, $d$, and that the FPFH features of each sampling point are distinct.

(2)

Find points in the target point cloud that have similar FPFH features to the sampling points in the source point cloud and consider them as corresponding points.

(3)

Determine the translation matrix and rotation matrix between the source and target point clouds based on the correspondence between them. Calculate the registration error using matrix transformation and estimate the error of the registration result using the Huber penalty function [24], denoted as ${\displaystyle \sum _{j=1}^{n}H(l_j})$. The expression is as follows:

$$H(l_j)=\left\{\begin{array}{cc}\frac{1}{2}{l}_j{}^2& ‖l_j‖\le m_l\\ \frac{1}{2}{m}_l(2‖l_j‖-m_l)& ‖l_j‖>m_l\end{array}\right.$$

(3)

where $m_l$ represents the predefined threshold; and $l_j$ represents the difference between the corresponding points in the $j$-th group.

(4)

Repeat the previous three steps until the registration error is minimized, obtaining the optimal transformation matrix to be used for subsequent fine registration.

2.3. Iterative Closest Point Fine Registration Algorithm

After the coarse registration of the slope stabilization structure point cloud, the source and target point clouds will have achieved a good initial alignment. Building upon the coarse registration, fine registration of the slope stabilization structure point cloud is performed to enhance the registration accuracy. The ICP algorithm is used for the fine registration, which iteratively obtains the optimal matching matrix through the least-squares method. The best transformation matrix is obtained when the Euclidean distance between the two point clouds is minimized. The process of the ICP fine registration algorithm is as follows [25]:

(1)

Find the nearest point set in the registration point cloud to the target point cloud, obtaining the initial correspondence between the two point clouds.

(2)

Based on the initial correspondence, compute the rotation matrix R and translation vector T between the two point clouds. Calculate the error using Equation (4).

$$F(R,T)=\frac{1}{N}{‖Q_i-(R\cdot P_i+T)‖}^2$$

(4)

where $Q_i$ represents the point in the target point cloud; and $P_i$ represents the point in the registration point cloud.

(3)

Set the threshold $\epsilon$ and maximum number of iterations $N_{\max }$. Update the position of the registration point cloud using the transformation matrix calculated in step (2) and obtain a new set of corresponding points $P_i{}^{\prime }$ using step (1).

(4)

Calculate the distance error between the target point cloud and the new set of corresponding points. If the iterative error is smaller than the threshold $\epsilon$ or the maximum number of iterations $N_{\max }$ is reached, terminate the iteration. Otherwise, go back to step (2) until the conditions are met.

3. Engineering Case Study

3.1. Overview of Slope Stabilization Structure Engineering

The research area of this paper is located on the southern section of Caihong Road in Changsha City, Hunan Province, China, spanning from the culvert under Yuelu Avenue to the north and ending at Yinshuang Road to the south. The supporting structure is an inclined gravity retaining wall. The original topography of the area is hilly with slight undulations, generally higher in the west and lower in the east. The ground elevation is approximately 41.20~61.10 m. The total length of the slope stabilization structure in the K0 + 220~K0 + 400 section (within the range of 220 m to 400 m from the starting point at 0 km of Caihong Road in the forward direction) of Golden Villa is about 170 m, with the maximum height at the top of the slope stabilization structure being 10 m and the slope stabilization structure being between 80~90°. The safety level of slope stabilization structure engineering is Grade I.

Additionally, the study area is located in a subtropical monsoon humid climate zone. According to meteorological department data, the annual average precipitation can reach 1394.6 mm, indicating significant rainfall. Therefore, it is essential to monitor the urban road slope stabilization structures to ensure their stability. If the deformation of the slope stabilization structure exceeds relevant standards, it is necessary to alert and take measures promptly to ensure slope stabilization structure stability. The schematic diagram of the location of the slope stabilization structure engineering area in the southern section of Caihong Road is shown in Figure 2.

3.2. Acquisition of Point Cloud Data

First, according to the specific morphology and terrain of the slope stabilization structure in the monitoring area, the locations and number of stations for setting up the scanner were chosen. Research by Guigang Shi et al. [26] on the median error under different numbers of station setups found that the number of station setups for point cloud registration should not exceed 4. To ensure the integrity of point cloud data, registration errors should be reduced, work efficiency should be improved, and the number of station setups should be minimized; three stations were set up in the monitoring area, named Station 1, Station 2, and Station 3. The point cloud data of the slope stabilization structure obtained using the Trimble SX10 device and the coordinate data of the feature points obtained using the Leica TCA2003 total station are shown in Figure 3.

In October 2021, the Trimble SX10 3D laser scanner was used to collect point cloud data from the slope stabilization structure in the southern section of Caihong Road. The basic parameters of the Trimble SX10 are shown in

Table 1. Performance parameters of Trimble SX10 3D laser scanner.

Instrument Name	Accuracy	The Maximum Range of EDM	Maximum Measuring Distance	Accuracy Standard	Maximum Scanning Rate
Trimble SX10	1″	800 m (with prism) 5500 m (without prism)	600 m	1 mm	26,600 pts/s

All values were taken from the technical specifications provided by the manufacturer.

. First, the 3D laser scanner was centered and leveled using the Trimble Access field book, selecting the “Station Setup” mode and entering the measured true height of the instrument into the field book. Second, to ensure a certain degree of precision in the point cloud registration, scanning was conducted using a polygonal field of view and a fine scanning density, specifically a scanning density of 2 mm@10 m, while ensuring that the overlap of the point cloud data between two adjacent stations was higher than 30%. Finally, after the parameter settings were completed, the slope stabilization structures at the three stations were scanned in sequence, taking a total of 1.2 h, with a total of 24,495,900 scanned point cloud datapoints. The slope stabilization structure point cloud data collected this time were taken as the first phase of the slope stabilization structure point cloud data.

Eight months later, in June 2022, a second scan of the slope stabilization structure was carried out. The steps for collecting the slope stabilization structure point cloud data were the same as in the first phase. The second scan took a total of 1.5 h, with a point cloud data count of 31,591,430.

3.3. Processing of Point Cloud Data

3.3.1. Registration of Point Clouds

The research area was composed of slope stabilization structure point cloud data scanned from three stations. Each station’s scanned point cloud data belonged to a different reference coordinate system, so it was necessary to register the point cloud data. This involved transforming the point cloud data from the three stations into the same coordinate system using a six-parameter rigid transformation. First, Station 1 was taken as the reference station, and Station 2 was the moving station. The geometric features of the slope stabilization structure point cloud region were extracted using the FPFH feature point extraction method. Next, the SAC-IA algorithm was used for the optimization and calculation to coarsely align the point cloud, providing it with a good initial position. Finally, the ICP algorithm was used to finely align the slope stabilization structure point cloud to improve the accuracy of the registration. After the registration was completed with the Trimble Business Center software, the residual value of the point cloud registration for Station 1 and Station 2 was 0.002 m, with an overlap of 39.32%. The combined slope stabilization structure point cloud of Station 1 and Station 2, after registration, was then taken as the reference station, and Station 3 was taken as the moving station. The registration process was the same as for Station 1 and Station 2, resulting in a registration residual value of 0.001 m and an overlap of 37.84%. The results of the slope stabilization structure point cloud before and after registration are shown in Figure 4. For general coordinate system registration, the accuracy requirement is 0~0.004 m. The error in this registration did not exceed 0.004 m, so it could meet the accuracy requirements of registration.

3.3.2. Denoising of Point Clouds

During the collection of the slope stabilization structure point cloud data with the Trimble SX10 3D laser scanner, some noise was scanned. This noise was caused by measurement errors in the laser scanning and included pedestrians, trees, vehicles, road signs, and telephone poles [27]. Research results show that the amount of noise in 3D laser scanning is about 0.1~5% of the total number of point clouds [28]. These noise points can affect the analysis of the deformation area of the slope stabilization structure and the calculation of deformation. Therefore, it was necessary to remove the noise points to ensure that the deformation monitoring results were not affected. First, the noise points far from the slope stabilization structure point cloud were manually removed. Second, the noise points near the slope stabilization structure point cloud were removed using point cloud segmentation, retaining the necessary point cloud data. After noise reduction, the number of slope stabilization structure point clouds decreased to 23,321,105. The results after noise reduction processing are shown in Figure 5.

3.3.3. Filtering of Point Clouds

Since the slope stabilization structure point cloud data were scanned through three measurement stations, there were certain overlapping areas in the registration process of the point cloud data from the different stations. This led to a high redundancy of point clouds, which affected the computational efficiency. In addition, the amount of point cloud data obtained by the 3D laser scanner was very large, which consumed a significant amount of system resources and affected the model reconstruction [29]. This resulted in very slow data processing. Therefore, by setting a threshold of 0.05 m to resample the point cloud, the number of slope stabilization structure point clouds was reduced from 23,321,105 after denoising to 8,955,304. The sampling rate was about 38.4%. After the filtering was complete, it not only retained the overall features of the slope stabilization structure but also reduced the storage space of the point cloud data, greatly improving the computational efficiency.

3.3.4. Reconstruction of Point Cloud Triangular Irregular Network Model

The Trimble Realworks software can reconstruct the surface of a slope stabilization structure point cloud data obtained from scanning, thereby analyzing the deformation between the two periods of slope stabilization structure point clouds. In the slope stabilization structure point cloud model, any point is arbitrarily selected, and the nearest point is found as the initial baseline of the irregular triangular network, which is constructed according to the Delaunay triangulation method [30]. The slope stabilization structure point cloud model is projected onto the computational analysis surface through surface projection, and the irregular triangular network is edited. This involves deleting protruding points in the irregular triangular network model and performing a second smoothing process, resulting in a complete slope stabilization structure model. The model of the slope stabilization structure point cloud irregular triangular network is shown in Figure 6.

3.4. Deformation Analysis

The point cloud data of the slope stabilization structure scanned in the second period were processed in the same way as the point cloud data of the slope stabilization structure in the first period, including registration, denoising, filtering, and reconstruction of the irregular triangular mesh model. By overlaying and projecting the irregular triangular mesh models of the slope stabilization structure point cloud from the two periods using the TRW and TBC software, the deformation of the slope stabilization structure was analyzed in terms of point deformation, overall deformation, and linear deformation.

3.4.1. Single-Point Deformation Analysis

Due to the lack of homonymous point information in the slope stabilization structure point cloud data scanned by the 3D laser scanner, it was necessary to find polygonal regions with easily identifiable feature points in the slope stabilization structure point cloud. The centroid method was used to determine the 3D coordinates of the centroid of the polygonal point cloud (this centroid served as a virtual feature monitoring point). A total of eight virtual feature monitoring points were identified, named P1, P2, P3, P4, P5, P6, P7, and P8, as shown in Figure 2. These eight virtual feature monitoring points were marked in the field using Leica reflector targets (50 × 50 mm), with the same numbering as above. The positions of the reflector targets should be stable and not easily movable. The point cloud processing flowchart for obtaining the virtual feature monitoring points of the slope stabilization structure using the centroid method is shown in Figure 7.

By using the centroid method, we obtained the coordinate data of the virtual feature monitoring points for the first and second phases of the Caihong Road slope stabilization structure. These were used as the first and second monitoring epochs for the analysis. After performing adjustment processing on the two monitoring epochs and synchronizing the two monitoring projects, we obtained the 3D displacement vector visualization plane (Figure 8) of the centroid of the polygon point cloud (virtual feature monitoring point). The values in the brackets in Figure 8 represent the 3D deformation values of a certain feature monitoring point with a unit of mm.

The direction indicated by the arrow is the direction of movement of the point cloud centroid, and the length of the arrow is the distance of the centroid’s movement. A longer arrow signifies a larger deformation value. As can be seen from Figure 8, among the eight points cloud centroids, the position of point P5 moved the most, with a maximum displacement of 8.52 mm. The position of point P6 moved the least, with a minimum displacement of 3.35 mm. The overall displacement trend of the slope stabilization structure was towards the southeast of the slope stabilization structure, and the displacement was relatively small. The 3D displacement vector visualization plane of the point cloud centroid could intuitively show the displacement situation of the centroid with a high degree of visualization.

In addition, using the Leica TCA2003 total station, eight feature monitoring points on-site were monitored once a month, and nine sets of monitoring data were obtained over an eight-month period. The displacement in the XYZ directions of the feature monitoring points is shown in Figure 9, and the performance parameters of the Leica TCA2003 total station are shown in

Table 2. Performance parameters of Leica TCA2003 total station.

Instrument Name	Angle Measurement Accuracy	Maximum Distance Measurement Accuracy	Search Accuracy
Leica TCA2003	0.5″	1 mm + 1 ppm	1 mm (range of 200 m)

All values were taken from the technical specifications provided by the manufacturer.

.

As can be seen from Figure 9, with time, the displacements in the XYZ directions of all feature monitoring points generally showed an increasing trend. The displacements of points P5 and P8 were the largest, with a maximum value of 6.2 mm. Furthermore, the displacement changes in the XYZ directions did not exceed 7 mm. After April 2022, the feature monitoring points tended towards a stable state, and the deformation results did not exceed the standards stipulated by regulations, indicating that the slope stabilization structure was safe and stable. As can be seen from Figure 8 and Figure 9, the displacement results of the feature monitoring points measured by the total station were consistent with the centroid displacement results obtained by the centroid method with small errors, indicating that the centroid method could be applied to point deformation analysis. However, since point deformation analysis can only analyze the deformation situation of specific points on a slope stabilization structure in a one-sided manner and cannot reflect the deformation situation of the local or entire slope stabilization structure, it was necessary to perform an overall deformation analysis and a line deformation analysis on the slope stabilization structure deformation.

3.4.2. Global Deformation Analysis

First, the slope stabilization structure point cloud obtained in the first phase was used as the base surface, and the slope stabilization structure point cloud obtained in the second phase was used as the comparison surface. According to the method of matching homonymous points, the triangulated irregular network models of the slope stabilization structure point clouds in the first and second phases were transformed and overlaid under the same coordinate system and projected in the manner of plane projection (YOZ plane). This resulted in a chromatic spectrum map (Figure 10) of the overall displacement deviation of the slope stabilization structure point cloud based on the same slope stabilization structure model for the two phases. The grid resolution of the chromatic spectrum map was set to 0.5 m.

As can be seen from Figure 10, the overall displacement deviation chromatic spectrum map of the Caihong Road slope stabilization structure appeared as blue–green from October 2021 to June 2022 over eight months. This indicated that the deformation of the entire slope stabilization structure area was relatively stable, with only a few areas in blue or yellow, indicating that the overall deformation range of the slope stabilization structure was controllable. Through interval statistical data calculation, the overall deformation of the slope stabilization structure ranged from −6.24 mm to 5.74 mm. Of this, 93.61% of the points had a deformation range between −0.76 mm and 0.92 mm, with a deformation of less than −0.76 mm accounting for 4.31% of the total number of points. The points where the slope stabilization structure deformation exceeded 0.92 mm accounted for 2.08% of the total number of points. The areas with significant displacement of the slope stabilization structure are indicated by the black box (areas with large negative deformation) and the red box (areas with large positive deformation), with deformation values not exceeding −6.24 mm and 5.74 mm, respectively. The statistical and Gaussian distribution graphs of the overall displacement deformation of the slope stabilization structure are shown in Figure 11.

3.4.3. Line Deformation Analysis

The deformation characteristics of single points could only reflect the deformation characteristics of the discrete points on the Caihong Road slope stabilization structure and could not reflect the continuous line deformation characteristics. Similarly, the point cloud overall displacement deviation chromatic spectrum could only reflect the deformation characteristics of the overall slope stabilization structure or local areas. Therefore, it was very necessary to perform a line deformation analysis on the deviation spectrum of the slope stabilization structure point cloud. As shown in Figure 10, the deformation was relatively large in the areas indicated by the black box and the red box on the slope stabilization structure. By setting horizontal and vertical sliders, the locations with significant deformation on the slope stabilization structure were cut with two horizontal and two vertical cuts (the red dashed lines from left to right represent the vertical slider 1 and the vertical slider 2, and the black dashed lines from top to bottom represent the horizontal slider 1 and the horizontal slider 2). The line chart of the face-to-face detection and the analysis of the two-phase slope point cloud triangular mesh model are shown in Figure 12. The displacement deviation chromatic spectrum of the horizontal and vertical sliders is shown in Figure 13, and the locations of the horizontal and vertical slider cut lines are shown in Figure 14.

In Figure 12, the red line represents the base plane section line of the first phase slope stabilization structure point cloud, and the green line represents the comparison plane section line of the second phase slope stabilization structure point cloud. As depicted in Figure 12, significant deformation occurred in the middle and left positions of the slope stabilization structure. Figure 14a illustrates that under the conditions of horizontal displacement slider 1, the slope stabilization structure experienced considerable deformation around 60~80 m, with a maximum deformation value of approximately 6.27 mm. Figure 14b shows that under the conditions of horizontal displacement slider 2, significant deformation happened at around 60~80 m, and the maximum deformation value was around 2.09 mm. Figure 14c indicates that under the conditions of vertical displacement slider 1, substantial deformation occurred at 6~10 m, with the maximum deformation value being less than 5.26 mm. Figure 14d demonstrates that under the conditions of vertical displacement slider 2, all the deformation values at 2~10 m were less than 2.08 mm. According to the Chinese industry standards “Technical Specification for Building Slope Engineering” (GB50330-2013), “Technical Specification for Monitoring of Building Foundation Pit Engineering” (GB50497-2016), and “Specification for Measurement of Building Deformation” (JGJ8-2016), the deformation values of the slope stabilization structure did not exceed the standards stipulated in these regulations. This suggests that the slope stabilization structure was stable and controllable from October 2021 to June 2022.

3.5. Measurement Accuracy Evaluation

To analyze the reliability of the Trimble SX10 3D laser scanner in measuring the slope stabilization structure deformation and to compare its measurement accuracy with that of the Leica TCA2003 total station, the centroid deformation obtained from the Trimble SX10 scan was evaluated for significant differences in accuracy against the deformation of the monitoring points measured by the Leica TCA2003 total station. The deviation values obtained from the 3D laser scanner and the total station were calculated using Equation (5). The results are shown in

Table 3. Deviation results statistics table.

Characteristic Monitoring Point	3D Laser Scanner Mode (mm)				Total Station Mode (mm)				Deformation Quantity $\left(\mathbf{m}\right) (△\mathit{d} = \mathit{d}-{\mathit{d}}^{\prime })$
	$\mathbf{\Delta }\mathit{x}$	$\mathbf{\Delta }\mathit{y}$	$\mathbf{\Delta }\mathit{z}$	$\mathit{d}$	$\mathbf{\Delta }{\mathit{x}}^{\prime }$	$\mathbf{\Delta }{\mathit{y}}^{\prime }$	$\mathbf{\Delta }{\mathit{z}}^{\prime }$	${\mathit{d}}^{\prime }$
P1	−3.61	3.04	−4.67	6.64	−3.47	3.23	−4.56	6.58	0.06
P2	−1.63	1.99	−2.85	3.84	−1.95	1.84	−2.64	3.76	0.08
P3	−2.92	1.97	−2.91	4.57	−2.61	2.33	−3.41	4.89	−0.32
P4	−2.67	2.94	−3.79	5.49	−2.87	2.85	−4.01	5.70	−0.21
P5	−4.08	4.34	−6.09	8.52	−4.21	4.12	−6.18	8.54	−0.02
P6	−2.11	1.43	−2.17	3.35	−1.82	1.77	−2.61	3.64	−0.29
P7	−2.88	2.76	−4.11	5.73	−3.01	2.79	−4.07	5.78	−0.05
P8	−3.74	3.91	−5.70	7.86	−3.99	3.87	−5.77	8.01	−0.15

:

$$\begin{gathered} d=\sqrt{\Delta x^2+\Delta y^2+\Delta z^2} \\ {d}^{\prime }=\sqrt{\Delta {x^{\prime }}^2+\Delta {y^{\prime }}^2+\Delta {z^{\prime }}^2} \end{gathered}$$

(5)

where $\Delta x$, $\Delta y$, $\Delta z$ represent the deviation in the $x$, $y$, and $z$ directions of the feature monitoring point in the 3D scanner mode, respectively; and $\Delta x$, $\Delta y$, $\Delta z$ represent the deviation in the $x$, $y$, and $z$ directions of the feature monitoring point in the total station mode, respectively.

The analysis was carried out using the method of an independent sample t-test, with $d_i,{d^{\prime }}_i(i=1,\cdots ,6)$ representing the sample values of the 3D laser scanner mode and the total station mode, respectively. Their sample difference is $\Delta d=d_i-{d^{\prime }}_i(i=1,\cdots ,6)$, and the test statistic $t$ can be expressed as:

$$t=\frac{\overline{\Delta d}-({\mu }_1-{\mu }_2)}{S/\sqrt{N}}$$

(6)

where $\overline{\Delta d}$ represents the mean value of $\Delta d$; ${\mu }_1,{\mu }_2$ represent the deviations of the 3D laser scanner and total station, respectively; $S$ represents the standard deviation of $\Delta d$; and $N$ represents the number of samples.

Assuming ${\mu }_1-{\mu }_2=0$, the statistic $t$ follows a t-distribution with N-1 degrees of freedom. The null hypothesis $H_0$ states that there was no significant difference between the slope stabilization structure measurement accuracy obtained from the 3D laser scanner mode and the total station mode. At the significance level $\alpha =0.05$, the test statistic t was calculated as 2.074, that is, $t<t_{0.025}(7)=2.365$, $P>0.05$. The results suggested that we accept the null hypothesis under the condition of a significance level of 0.05, meaning that the measurement accuracy of the 3D laser scanner mode and the total station mode was highly consistent and could satisfy the precision requirements for slope stabilization structure deformation. This demonstrates that the precision of the 3D laser scanner applied to slope stabilization structure deformation analysis is reliable.

4. Conclusions

This study analyzed the deformation characteristics of the two phases of the Caihong Road slope stabilization structure using ground 3D laser scanning technology. Based on the analysis results, the following conclusions can be drawn:

(1)

The Trimble SX10 scanning technology, with its advantages of being non-contact, having a rapid scanning speed, and having a high work efficiency, avoids the limitations of traditional monitoring methods that are point-based. It carries significant implications for slope stabilization structure deformation analysis.

(2)

The centroid displacement results obtained by the centroid method were in good agreement with the displacement results of the feature monitoring points measured by the total station, with small errors. Furthermore, the maximum displacement of the centroid method did not exceed 9 mm, validating the suitability of the centroid method for slope stabilization structure point deformation analysis.

(3)

In the analysis of the overall deformation and line deformation, the line deformation could provide a detailed and accurate analysis of all or part of the overall deformation. The overall slope stabilization structure point cloud displacement deviation spectrum showed that 93.61% of the point deformation range was between −0.76~0.92 mm. The line deformation analysis showed that the slope stabilization structure deformation did not exceed 6.27 mm for both the horizontal and vertical displacement lines, both of which did not exceed the standard specified by the regulations. This indicates that the slope stabilization structure was in a safe and reliable state, and the line deformation analysis could supplement the overall deformation analysis.

(4)

The independent sample t-test method further verified the reliability of the measurement accuracy of the 3D laser scanning technology in measuring the slope stabilization structure deformation. The 3D laser scanning technology could also provide theoretical and technical references for similar practical projects such as building settlements, tunnel deformation, and bridge displacement.

The application of 3D laser scanning technology to monitor the deformation of slopes has several contributions to society. First, it can provide accurate and objective data on the deformation of slopes, which can be used to assess the safety and stability of slopes. This is particularly important for areas prone to natural disasters such as landslides, where early warning systems can be put in place based on the data obtained through 3D laser scanning. Additionally, it can reduce the time and cost required to monitor slopes compared to traditional methods. This allows for more frequent and regular monitoring, improving the accuracy and reliability of the data obtained. However, there are certain limitations to monitoring slope deformation using 3D laser scanning technology. For example, obstacles on the surface of the slope and uneven terrain can affect the accuracy and reliability of the scans. Nevertheless, the future research prospects for monitoring slope deformation using 3D laser scanning technology are promising. The technology can be combined with remote sensing techniques, drone surveys, and other methods to further improve the accuracy and reliability of slope deformation monitoring. Additionally, future research may focus on the predicting and warning of slope instability, as well as conducting comprehensive assessments and analyses of slope stability.