Application of deformation detection method based on time-varying point cloud in the protection of ancient buildings

Abstract

In order to detect the deformation of ancient buildings completely, accurately and continuously, a time-varying point cloud group is established by using three-dimensional laser scanner to detect and sample ancient buildings in different time periods. By comparing and calculating time-varying point clouds, the deformation of ancient buildings can be comprehensively analysed. In order to improve the efficiency of data processing, a BSP parallel algorithm for deformation analysis based on time-varying point clouds is designed, and the processing and calculation of deformation data are completed by a computing cluster composed of several independent computing units. This method of deformation detection based on time-varying point cloud has been applied to many ancient building detection and protection projects such as Hailongdun site in Zunyi, Guizhou Province China, Huangze Temple in Guangyuan, Sichuan Province China, etc. The application results show that the deformation monitoring method based on time-varying point cloud is quicker, more economical, more accurate and less harmful to ancient buildings than the traditional method. Compared with the general point cloud deformation analysis methods, the accuracy and comprehensiveness of the deformation monitoring results are higher. It can be used as one of the best methods for deformation monitoring of ancient buildings.

Keywords

Ancient building deformation detection point cloud parallel computing

1. Introduction

Geometric data information of surface of objects collected by three-dimensional laser scanning system is called point cloud. By processing and calculating the point cloud data and analysing the deformation of objects, it is a new deformation detection technology [1]. It has been widely used in the fields of laser remote sensing measurement, engineering and environmental detection, and has great application prospects in the field of ancient building survey and protection [2, 3].

Traditional deformation detection methods of ancient buildings mostly use common measuring tools, rangefinders, total station and so on to calculate the deformation of ancient buildings by measuring the position difference of a few specific points in different periods [4]. Compared with the traditional methods, the point cloud-based deformation detection technology using three-dimensional laser scanner is fast, economical, accurate, and has no contact damage to ancient buildings. It is one of the ideal methods for deformation detection and protection of ancient buildings [5, 6].

Generally, deformation analysis based on point clouds is mostly carried out by comparing the distances of two-point clouds. In fact, due to the complexity of the ancient building form, the difference of point cloud collection environment, and the limitation of the accuracy of collection equipment, the collected point cloud data are very different. Based on only two or a few scans, the deformation or deformation domain obtained by comparing between two-point clouds is usually unstable or irrelevant. Based on this unstable or irrelevant calculation results, it is difficult to calculate the shape changes of ancient buildings, at least inaccurate and comprehensive [7, 8]. In order to detect the shape change of ancient buildings completely, accurately and continuously, a time-varying point cloud group can be established by sampling the ancient buildings in different time periods. By comparing and calculating the time-varying point clouds, the shape change of ancient buildings can be analysed comprehensively. Based on the deformation analysis of time-varying point clouds, compared with the deformation method calculated only by comparing two-point clouds, it cannot only measure the deformation of ancient buildings more accurately, but also measure the area where the deformation occurs, and further analyse the deformation characteristics and trends of ancient buildings. Its accuracy and comprehensive analysis performance are stronger.

The deformation analysis based on time-varying point clouds, it will involve parallel processing and calculation of multi-point clouds. In order to improve the processing efficiency, we use Bulk Synchronous Parallel (BSP) Computing Technology [9] to construct the algorithm. Through a computing cluster composed of several independent computing units [10, 11], we complete the parallel processing calculate of the deformation analysis data of time-varying point clouds.

The proposed deformation detection method based on time-varying point clouds has been applied to many deformation detection and protection projects of ancient buildings, and its characteristics of fast, economical, little contact damage to ancient buildings, and high accuracy and comprehensiveness of deformation detection have been fully verified. In this paper, the typical application examples of time-varying point cloud-based deformation detection method in Hailongdun site in Zunyi, Guizhou Province China, and Huangze Temple in Guangyuan, Sichuan Province China are listed for reference.

2. Deformation calculation

At $t_{0}$ , the collected point cloud is $P_{0}$ , which serves as a reference point cloud for deformation analysis. In a detection period $[t_{1},t_{m}]$ , $m$ times of detection sampling are carried out, and the point clouds obtained are used as the contrast point clouds for deformation analysis, denoted as:

sampling time: $\ t_{1},t_{2},\cdots,t_{i},\cdots,t_{m}$ ,

contrast point clouds: $P_{1},{P}_{2},\cdots,P_{i},\cdots,P_{m}$ .

Before deformation analysis, point clouds $P_{0},P_{1},P_{2},\ldots,P_{i},\ldots,P_{m}$ are processed and registered, and point clouds $P^{(0)},P^{(1)},P^{(2)},\ldots,P^{(i)},\ldots,P^{(m)}$ and registration errors ${\varepsilon}_{0},{\varepsilon}_{1},{\varepsilon}_{2},\ldots,{\varepsilon}_{i}% ,\ldots,{\varepsilon}_{m}$ are obtained. Among them, ${\varepsilon}_{0}$ represents the error of three-dimensional laser scanner, ${\varepsilon}_{i},i\neq 0$ represents the registration error of point cloud $P^{(i)}$ at time $t_{i}$ to point cloud $P^{(i-1)}$ at time $t_{i-1}$ [12].

The shortest distance $d$ (Harsdorf distance) from any point $p^{(i)}$ in point cloud $P^{(i)}$ to point cloud $P^{(0)}$ can be calculated by Eq. (1).

$\displaystyle d\left(p^{\left(i\right)},P^{\left(0\right)}\right)={\textit{min% }}_{p^{(0)}\in P^{(0)}}d||p^{(i)}-p^{(0)}||_{2}$ (1)

If $d(p^{(i)},P^{(0)})\neq 0$ , then $p^{(i)}$ is considered to be a deformed point. The set of all deformed points in $P^{(i)}$ forms a set of deformed points, which is denoted as $\partial P^{(i)}$ . Deformation point $p^{(i)}$ corresponds to the point in the reference point cloud $P^{(0)}$ and its adjacent points, which are called deformational reference points. The set of all deformational reference points in $P^{(0)}$ forms a set of deformational reference points, which is denoted as $\partial P^{(0)}$ . The irregular cube surrounded by point set $\partial P^{(0)}$ and $\partial P^{(i)}$ is called the deformed body of point cloud $P^{(i)}$ in $t_{i}$ period as compared with reference point cloud $P^{(0)}$ . It is denoted as $\Delta S^{(i)}=[\partial P^{(i)},\partial P^{(0)}]$ , as shown in Fig. 1.

Figure 1.

Schematic diagram of the deformed body $\Delta S^{(i)}=[\partial P^{(i)},\partial P^{(0)}]$ .

For a continuously deformed object, the deformed bodies in different time periods are different. There may be more than one deformed body of a detected object, that is, the deformation does not necessarily occur at every point of the detected object. In order to accurately analyse the deformation of the detected object, each deformed body should be analysed one by one.

The maximum deformation of the deformed body is estimated by the maximum difference of the components of the deformed body in the X, Y and Z directions. Among them, the maximum deformation of the deformed body in the X direction is $\Delta x^{(i)}_{\textit{max}}=|x^{(i)}_{\textit{max}}-x^{(i)}_{\textit{min}}|$ and the maximum deformation in the Y direction is $\Delta y^{(i)}_{\textit{max}}=|y^{(i)}_{\textit{max}}-y^{(i)}_{\textit{min}}|$ and the maximum deformation in the Z direction is $\Delta z^{(i)}_{\textit{max}}=|z^{(i)}_{\textit{max}}-z^{(i)}_{\textit{min}}|$ .

The irregular polygon formed by the connection of the boundary projection points of the deformed body $\Delta S^{(i)}$ on the $X\_Y$ plane is called the deformed region of the deformed body $\Delta S^{(i)}$ on the $X\_Y$ plane and is denoted as $\Delta S^{(i)}_{X\_Y}i\in(1,2,\ldots,m)$ . Similarly, the deformation domain of $\Delta S^{(i)}$ on the $Y\_Z$ plane is $\Delta S^{(i)}_{Y\_Z}\ i\in(1,2,\ldots,m)$ , and the deformation domain of $\Delta S^{(i)}$ on the $X\_Z$ plane is $\Delta S^{(i)}_{X\_Z}\ i\in(1,2,\ldots,m)$ .

For the deformed point set $\partial P^{(i)}$ in $P^{(i)}$ , one point $p^{(i)}_{i}$ and two adjacent points $p^{(i)}_{j}$ , $p^{(i)}_{k}$ are connected to form a plane triangle. According to Delaunay’s spatial region growth algorithm, the deformed point set $\partial P^{(i)}$ is triangulated to form a triangular network surface. Similarly, the deformation reference point set $\partial P^{(0)}$ is triangulated to form a $\partial P^{(0)}$ triangular network surface. The size of the projection area of the deformed body $\Delta S^{(i)}$ on the $X\_Y$ plane, i.e. the area of the deformed body $\Delta S^{(i)}_{X-Y}$ is $\Delta A^{(i)}_{X-Y}$ :

$\displaystyle\Delta A^{(i)}_{X-Y}=\Sigma{\partial}A^{(i)}_{ijk},\Delta A^{(0)}% _{X-Y}=\Sigma{\partial}A^{(0)}_{ijk}$ (2)

In Eq. (2), $\partial A^{(i)}_{ijk},\partial A^{(0)}_{ijk}$ are the projection areas of the two triangles on the $X\_Y$ plane, respectively. These two triangles are composed of three points in $\partial S^{(i)}$ and $\ \partial S^{(0)}$ , $p^{(i)}_{i},p^{(i)}_{j}p^{(i)}_{k}$ and $p^{(0)}_{i},p^{(0)}_{j}.p^{(0)}_{k}$ . Similarly, the projection areas of the deformed domains $\Delta S^{(i)}_{X-Z}$ , $\Delta S^{(i)}_{Y-Z}$ and $\Delta A^{(i)}_{X-Z}$ and $\Delta A^{(i)}_{Y-Z}$ can be obtained.

3. Algorithm design

The deformation analysis and calculation of time-varying point clouds can be realized by a parallel computing cluster composed of $m+1$ independent computing units [13]. The computing cluster is located in a LAN segment and its topology is shown in Fig. 2. Each independent computing unit corresponds to a computing node, which can be a common PC. The size of the computing cluster can be determined by the size of the computing task. Generally, $m\geqslant 3$ .

In a computing cluster, one of the computing units is set as the control entry, which is responsible for the management of file system namespace, assignment and boundary synchronization of computing tasks of each computing node, and is denoted as $U_{0}$ . The other computing units are computing nodes, which are responsible for data storage, processing and calculation of point clouds, and are denoted as $U_{1},U_{2},\ldots,U_{m}$ .

Figure 2.

The topology of computing cluster ( $1\leqslant i\leqslant m$ ).

The point cloud $P^{(0)},P^{(1)},P^{(2)},\ldots,P^{(i)},\ldots,P^{(m)}$ are allocated to the computing node $U_{0},U_{1},U_{2},$ $\ldots,U_{m}$ correspondingly. Among them, $U_{0}$ is set as the computing node of $P^{(0)}$ of reference point cloud and serves as the computing node of comprehensive data processing.

According to Bulk Synchronous Parallel (BSP) Computing Technology, the deformation analysis calculation based on time-varying point cloud will be carried out in the following six super-steps [14, 15].

•

Super Step 1

Global Communication 1-1: $U_{0}$ node sends the address of reference point cloud $P^{(0)}$ in the computing cluster system and ${\varepsilon}_{0}$ to other nodes, which can be used to compare the deformation calculation of point cloud and reference point cloud.

•

Super Step 2

Local Computation 2-1: At $U_{i},\ i\in\{1,2,\ldots,m\}$ nodes, compute the set of k-neighbourhood points $P^{(0)}_{k}$ of point $p^{(i)}$ ( $\in P^{(i)}$ ) in reference point cloud $P^{(0)}$ . According to the least square plane fitting algorithm, the local plane ${\Sigma}^{(0)}_{k}$ is obtained, and the distance $d^{(i)}$ from point $p^{(i)}$ to local plane ${\Sigma}^{(0)}_{k}$ is calculated.

Global Communication 2-2: $U_{i},\ i\in\{1,2,\ldots,m\}$ nodes send deformation calculation reports to $U_{0}$ .

•

Super Step 3

Local Computation 3-1: $U_{0}$ node analyse the deformation situation: if the deformation is not judged, the algorithm ends, and gives the final report of the comparison results including all point clouds; if the deformation is judged, the deformation occurs at the time $\{t_{d(1)},t_{d(2)},\ldots,t_{d(q)}\}\subseteq\{t_{1},t_{2},\ldots,t_{i},% \ldots,t_{m}\}$ , according to the deformation occurrence time $\{t_{d(1)},t_{d(2)},\ldots,t_{d(q)}\}$ , deployment follow-up algorithm.

Global Communication 3-2: $U_{0}$ nodes send deformation analysis instructions to the corresponding computing nodes at the time of occurrence $\{t_{d(1)},t_{d(2)},\ldots,t_{d(q)}\}$ .

•

Super Step 4

Global Communication 4-1: The point cloud $P^{d(i)}$ and registration error ${\varepsilon}_{d(i)}$ are sent by $U_{d(i)}$ node to $U_{d(i-1)}$ node, where $i\in\{1,2,\ldots,q\}$ . It is used to analyze the deformation trend of point cloud data at two adjacent time points.

•

Super Step 5

Local Computation 5-1: At $U_{d(i)},i\in\{1,2,\ldots q\}$ node, for point $p^{d(i)}$ in point cloud $P^{d(i)}$ , the k-neighbourhood point set $P^{d(i-1)}_{k}$ of point $p^{d(i)}$ in point cloud $P^{d(i-1)}$ at time $t_{d(i-1)}$ is calculated. The local plane ${\Sigma}^{d(i-1)}_{k}$ is obtained by least squares plane fitting algorithm, and the distance $d^{d(i)}$ from point $p^{d(i)}$ to local plane ${\Sigma}^{d(i-1)}_{k}$ is calculated. Let the threshold value of basic error $\varepsilon={\varepsilon}_{0}+{\varepsilon}_{d(i-1)}$ . If the distance $d^{d(i)}$ exceeds the basic error threshold $\varepsilon$ , the point $p^{d(i)}$ is regarded as a deformed point.

Local Computation 5-2: At $U_{d(i)},i\in\{1,2,\ldots q\}$ node, the set of all deformed points in $P^{d(i)}$ is composed of a set of deformed points, which is denoted as $\delta P^{d(i)}$ . At $t_{d(i-1)}$ moment, the point in the point cloud $P^{d(i-1)}$ corresponding to the deformation point $p^{d(i)}$ is taken as the deformation reference point. The set of all deformation reference points in $P^{d(i-1)}$ is composed of a set of deformation reference points, which is denoted as $\delta P^{d(i-1)}$ .

•

Super Step 6

Local Computation 6-1: At $U_{d(i)},\ i\in\{1,2,\ldots,m\}$ node, feature extraction is carried out for deformation point set $\delta P^{d(i)}$ and deformation reference point set $\delta P^{d(i-1)}$ respectively. For the set of deformation feature points and the set of deformation reference feature points, the one-to-one correspondence between the two sets of feature points is analysed, and the relation of deformation point $C^{d(i-1)\Rightarrow d(i)}$ is obtained.

Global Communication 6-2: $U_{d(i)}$ node sends the deformed point relation $C^{d(i-1)\Rightarrow d(i)}$ to $U_{m}$ , which serves as the trend of $P^{(m)}$ of the nearest time $t_{m}$ in the detection period $[t_{1},t_{m}]$ .

The results of deformation analysis are visualized according to two categories: (1) the degree of deformation of point cloud data $P^{(i)}$ at time $t_{i}$ and point cloud data $P^{(0)}$ at reference time $t_{0}$ , $i\neq 0$ ; (2) the point cloud data $P^{(i)}$ at time $t_{i}$ and point cloud data $P^{(i-1)}$ at last time $t_{i-1}$ , $i\neq 0$ .

4. Application analysis

4.1 Deformation analysis of tiezhu-gate in hailongdun site

Located in Zunyi, Guizhou Province China, the World Cultural Heritage Site of Hailongdun is a chieftain Site which combines large military buildings with palace buildings. It is also a well-preserved Castle site of Ming Dynasty. It has a history of more than 400 years.

Since January 2015, we have carried out continuous detection, sampling and deformation analysis of Hailongduan sites. Since most of the sites have been repaired or rebuilt, many high-risk buildings have been strengthened and no significant deformation has been detected so far. However, some sites still show signs of deformation, such as the Tiezhu-Gate in Hailongdun Site, due to the construction of surrounding projects and the impact of a large number of tourists.

In order to investigate the impact of engineering construction and tourist tours on the Tiezhu-Gate in Hailongdun site, we have been monitoring and sampling the Tiezhu-Gate in Hailongdun site for nearly two years since March 2015. The laser scanning equipment used in the detection and sampling is Z $+$ F 5010C IMAGE, and a number of point clouds on the top of the Tiezhu-Gate in Hailongdun site are obtained. According to the above algorithm, point clouds after basic processing and registration calculation are formed into a group of time-varying point clouds, which are loaded into BSP computing cluster. After analysis and calculation, point clouds $P^{(0)}$ , $P^{(1)}$ , $P^{(2)}$ (collected in 2015/3/27, 2016/5/16 and 2017/3/7, respectively) with relatively stable deformation and highly correlated location in deformation domain are obtained, as shown in Fig. 3.

Figure 3.

Point cloud $P^{(0)}$ , $P^{(1)}$ , $P^{(2)}$ of the Tiezhu-Gate.

Figure 4.

Deformation comparison of $P^{(1)}$ with $P^{(0)}$ .

For point clouds $P^{(0)}$ , $P^{(1)}$ , $P^{(2)}$ , the results of deformation analysis are as follows: through processing and calculating the point cloud data, the registration error from point cloud $P^{(1)}$ registration to $P^{(0)}$ is ${\varepsilon}_{0\leftarrow 1}=$ 0.100 mm, the registration error from point cloud $P^{(2)}$ registration to $P^{(0)}$ is ${\varepsilon}_{0\leftarrow 2}=$ 0.100 mm, and the acquisition error of Z $+$ F 5010C IMAGE device is ${\varepsilon}_{0}=$ 0.200 mm. According to the above algorithm, the deformation analysis and calculation are carried out, the comparison results of point cloud $P^{(1)}$ with $P^{(0)}$ are obtained, as shown in Fig. 4, and the comparison results of point cloud $P^{(2)}$ with $P^{(0)}$ are obtained, as shown in Fig. 5.

Figure 5.

Deformation comparison of $P^{(2)}$ with $P^{(0)}$ .

Figure 6.

Point cloud $P^{(0)},P^{(1)},P^{(2)}$ of the north wall of Feifeng-Gate.

As shown in Figs 4 and 5, the maximum deformation of point cloud $P^{(1)}$ relative to $P^{(0)}$ is about 0.150 mm. The maximum deformation of point cloud $P^{(2)}$ relative to $P^{(0)}$ is about 0.300 mm. The detected deformation is small and concentrated in the front and middle part of the Tiezhu-Gate in Hailongdun site.

It can be considered that the surrounding engineering construction and tourists’ visits have a certain inherent impact on the stability of the Tiezhu-Gate in Hailongdun site, but the impact is small, so we should pay attention to strengthening protection.

Figure 7.

Deformation comparison of $P^{(1)}$ with $P^{(0)}$ .

Figure 8.

Deformation comparison of $P^{(2)}$ with $P^{(0)}$ .

4.2 Deformation detection of Feifeng-gate in hailongdun site

In Hailongcun site, there are 9 the Gates, including Chaotian-Gate, Tiezhu-Gate and Feifeng-Gate and so on. Due to the long-term weather erosion and sunshine exposure, the walls of the Gates cracked in varying degrees, of which the northern wall of the Feifeng-Gate was the most serious.

In order to detect the change of wall cracks of Feifeng-Gate, we continuously sampled the north wall of Feifeng-Gate from March 2015. Optech ILRIS-3D was used to collect data in March 2015, and Z $+$ F 5010C IMAGE is used in the data acquisition in the future. Similarly, according to the above algorithm, after basic processing and registration calculation, a group of time-varying point clouds is formed for the scanning point clouds of the northern wall of Feifeng-Gate, which are loaded into the BSP computing cluster. After analysis and calculation, the point clouds $P^{(0)}$ , $P^{(1)}$ , $P^{(2)}$ (collected in 2015/3/25, 2016/5/17 and 2017/3/10, respectively) with relatively stable deformation and highly correlated location in the deformation domain are obtained, as shown in Fig. 6.

Figure 9.

Point cloud $P^{(1)}$ of Yinghuilou Buddha.

Figure 10.

Point cloud $P^{(2)}$ of Yinghuilou Buddha.

For point clouds $P^{(0)}$ , $P^{(1)}$ , $P^{(2)}$ , the results of deformation analysis are as follows: through processing and calculating the point cloud data, the registration error of point cloud $P^{(1)}$ to $P^{(0)}$ is ${\varepsilon}_{0\leftarrow 1}=$ 0.615 mm; the registration error of point cloud $P^{(2)}$ to $P^{(0)}$ is ${\varepsilon}_{0\leftarrow 2}=$ 0.573 mm; the registration error of point cloud $P^{(2)}$ to $P^{(1)}$ is ${\varepsilon}_{1\leftarrow 2}=$ 0.184 mm. In addition, the acquisition error of Optech ILRIS-3D device is ${\varepsilon}^{1}_{0}=$ 0.500 mm, and that of Z $+$ F 5010C IMAGE device is ${\varepsilon}^{2}_{0}=$ 0.200 mm. Comparing the deformation of point cloud $P^{(1)}$ with $P^{(0)}$ , as shown in Fig. 7, and the deformation of point cloud $P^{(2)}$ with $P^{(0)}$ , as shown in Fig. 8.

Figure 11.

Deformation comparison of $P^{(2)}$ with $P^{(1)}$ .

Figure 12.

Histogram of deformation distribution.

As shown in Figs 7 and 8, the deformation of point cloud $P^{(1)}$ and $P^{(2)}$ with $P^{(0)}$ is smaller. The maximum deformation is only 0.020 mm, which is less than the equipment error and registration error. It can be concluded that the cracks in the north wall of Feifeng-Gate in Hailongdun Site have not changed during the detection period.

4.3 Weathering detection of yinghuilou buddha statues of huangze temple

Huangze Temple in Guangyuan, Sichuan Province China, was built more than 1300 years ago. The 41 Buddha statues in the grottoes were carved on a sandstone cliff. Because of long-term weather erosion and sunshine exposure, the grotto Buddha statues have shown different degrees of weathering erosion. At the request of Huangze Temple Museum, since 2015, we have been digitally collecting all the grotto Buddha statues, at the same time, we have been detected the weathering of several key protected grotto Buddha statues represented by Yinghuilou Buddha statues.

Figure 9 shows the point cloud $P^{(1)}$ of the Yinghuilou Buddha statue collected by the Huangze Temple Museum in July 2015. The measurement error of the acquisition equipment is ${\varepsilon}^{1}_{0}=$ 0.200 mm. Figure 10 shows the point cloud data $P^{(2)}$ collected in September 2016. The device used is the REVScan hand-held scanner, and the acquisition error is ${\varepsilon}^{2}_{0}=$ 0.200 mm.

By processing and calculating the point cloud data, the registration error of point cloud $P^{(2)}$ to $P^{(1)}$ is ${\varepsilon}_{1\leftarrow 2}=$ 0.752 mm. According to the above algorithm, the deformation comparison of $P^{(2)}$ with $P^{(1)}$ and the histogram of deformation distribution are obtained, as shown in Figs 11 and 12.

As shown in Figs 11 and 12, the Buddha statue of Yinghuilou Grottoes has a slight deformation within a one-year detection period, with the maximum deformation less than 0.042 mm. Due to the limitation of acquisition accuracy and the unavoidable acquisition noise, the deformation area can only be observed roughly from the highlighted area in Fig. 11.

The above results of deformation detection and analysis are consistent with the field observation results of cultural relics protection experts, and have been confirmed by the relevant cultural relics departments.

The application results show that for Hailongdun Site and Huangze Temple ruins, which are located in the alpine field and have a long history of earth-rock structure, the collected point cloud data are quite different due to the complexity of the surface morphology of the buildings, the change of the point cloud collection environment and the limitation of the accuracy of the collection equipment. The general point cloud deformation analysis method, based on two or a few scans of data, compares and calculates between two-point clouds. The deformation or deformation domain obtained is usually unstable or irrelevant. Based on this unstable or irrelevant calculation results, it is difficult to calculate the shape changes of ancient buildings. Based on the time-varying point cloud deformation detection method, point cloud data are collected in a planned time-varying period, and the acquisition environment is as close as possible. After basic processing and registration calculation, point clouds form a group of time-varying point clouds, which are loaded into parallel computing cluster. By comparison and comprehensive analysis, the deformation domain and the stable deformation correlation point clouds can be obtained. Through the analysis of the deformation correlation point clouds, not only the deformation quantity of ancient buildings can be obtained, but also the area where the deformation occurs can be measured and further divided. The deformation characteristics and trends of ancient buildings are analysed. That is to say, in terms of accuracy and comprehensiveness of deformation monitoring of ancient buildings, the deformation detection method based on time-varying point cloud is obviously superior to the general point cloud deformation analysis method,

5. Conclusion

By processing and calculating the point cloud data collected by the three-dimensional laser scanning system, the deformation of objects is analysed, which is a new type of deformation detection technology. Compared with traditional building deformation detection methods such as common measuring tools, range finders, total station, etc., the deformation detection technology based on point cloud is fast, economical, accurate, and has little contact damage to ancient buildings. It is one of the ideal methods for deformation detection and protection of ancient buildings.

Generally, the point cloud deformation analysis is mostly carried out by comparing the distances between two-point clouds, which is difficult to accurately and comprehensively analyse the deformation of ancient buildings. In order to detect the deformation of ancient buildings completely, accurately and continuously, three-dimensional laser scanner can be used to detect and sample the ancient buildings in different time periods, and a time-varying point cloud group can be established. By comparing and calculating the time-varying point clouds, the deformation of ancient buildings can be analysed comprehensively. In order to improve the efficiency of data processing, BSP parallel algorithm for time-varying point cloud deformation analysis is designed, and a computing cluster composed of several independent computing units is used to complete the processing and calculation of its deformation data. The proposed deformation detection method based on time-varying point clouds has been applied to many deformation detection and protection projects of ancient buildings, such as Hailongdun Site in Zunyi, Guizhou Province China, Huangze Temple in Guangyuan, Sichuan Province China, etc. The application results show that the deformation monitoring method of ancient buildings based on time-varying point clouds is faster, more economical, more accurate and less harmful to ancient buildings than the traditional method of deformation monitoring of ancient buildings. Compared with the general point cloud deformation analysis methods, it can not only measure the deformation of ancient buildings more accurately, but also measure the area where the deformation occurs, and the deformation characteristics and trends of ancient buildings are analysed. Its accuracy and comprehensive analysis performance are stronger. It is one of the best methods for deformation monitoring of ancient buildings.

Footnotes

Acknowledgments

This paper is the research result of the project “Research and Implementation of a Mobile Multi-view Registration System” funded by the Doctoral Research Fund of Guizhou Normal University (BoKeZi [11904-0517081]).

References

Tsakiri

and Anagnostopoulos

, Change detection in terrestrial laser scanner data via point cloud correspondence, International Journal of Engineering Innovation & Research (6) (2015), 125–131.

Fanti

Gigli

Lombardi

Tapete

and Canuti

, Terrestrial laser scanning for rockfall stability analysis in the cultural heritage site of Pitigliano (Italy), Landslides (4) (2013), 77–84.

Sun

Jiang

Zhou

and Weng

, The method of building multi-mode fusion ancient architecture model, Laser and Infrared 49(3) (2019), 296–302.

and Du

, Application of three-dimensional laser scanning technology in tomb surveying and mapping, Surveying and Mapping Bulletin (1) (2017), 161–162.

Sun

Wang

and Liu

, Application of three-dimensional laser scanning in surveying and mapping of ancient buildings, Surveying and Mapping Science 41(12) (2016), 297–301.

Ding

Yan

and Lu

, Application of ground three-dimensional laser scanning technology in deformation monitoring of ancient buildings, Mine Survey 45(5) (2017), 19–22.

Chen

Deng

Fan

and Hao

, Large-scale deformation monitoring in mining area by D-InSAR and 3D laser scanning technology integration, International Journal of Mining Science and Technology (4) (2013), 46–53.

Walton

Delaloye

and Diederichs

M.S.

, Development of an elliptical fitting algorithm to improve change detection capabilities with applications for deformation monitoring in circular tunnels and shafts, Tunnelling and Underground Space Technology (5) (2014), 178–185.

Bao

Wang

et al., BC-BSP: a BSP-based parallel iterative processing system for big data on cloud architecture, Proceedings of the 18th International Conference on Database Systems for Advanced Applications, 2013, 327–334.

10.

Chen

Song

and Yang

, Based on Hama parallel ant colony algorithm model and TSP application research, Transportation System Engineering and Information 16(3) (2016), 168–173.

11.

Liu

Zhang

Yuan

and Liu

, Fast nonlinear iterative method for power flow calculation based on graph computing, Power Information and Communication Technology 16(10) (2018), 19–24.

12.

Zhu

and Li

, Continuous extraction and deformation analysis method of tunnel section based on laser point cloud, Journal of Surveying and Mapping (9) (2015), 89–97.

13.

Liu

and Wang

, Research and implementation of parallel programming model in virtual computing environment, Microelectronics and Computer 32(12) (2015), 141–146, 150.

14.

Khayyat

Awara

Alonazi

et al., Mizan: a system for dynamic load balancing in large-scale graph processing, The 8th ACM European Conference on Computer Systems, 2013, pp. 275–280.

15.

Travelletti

Malet

J.-P.

and Delacourt

, Image-based correlation of laser scanning point cloud time series for landslide monitoring, International Journal of Applied Earth Observations and Geoinformation (10) (2014), 207–213.