4D POINT CLOUD (4DPC)-DRIVEN REAL-TIME MONITORING OF CONSTRUCTION
MOBILE CRANES
Dong Liang1, Zhe Chen1, Lingming Kong1, Yijie Wu1, Sou-Han Chen1, and Fan Xue1
1
The University of Hong Kong, Hong Kong SAR, China
This is the post-print version of the paper:
Liang, D., Chen, Z., Kong, L., Wu, Y., Chen, S.-H., & Xue, F. (2023). 4D point cloud (4DPC)driven real-time monitoring of construction mobile cranes. In Proceedings of 2023 European
Conference on Computing in Construction cum the 40th International CIB W78 Conference, 258.
Heraklion, Crete, Greece. Doi: https://doi.org/10.35490/EC3.2023.258

This file is released under the CC-BY NC 4.0 license, any other use of this file must seek for
prior permission approval by the authors.

Abstract
Mobile cranes are essential equipment in construction
sites due to their high flexibility and mobility. However,
the existing sensing or monitoring methods have
limitations in monitoring mobile cranes on sites.
Recently, the advent of 4D point cloud (4DPC)
technology with a unique spatial-temporal data structure
has shown potential in addressing these issues. In this
paper, we present a 4DPC monitoring approach, which
includes a set of prototype devices and a rule-based object
detection method. We conducted a proof-of-concept test
to monitor the hoisting process of two H-beams in a
footbridge construction project. The rule-based object
detection method successfully detected the target beams
in the collected six-hour 4DPC data. In the future, we
expect more efficient and robust 4DPC sensing devices
and processing methods for proactive crane motion
prediction and optimization in a time-dynamic site
environment.

1 Introduction
Cranes are critical equipment for hoisting and
transporting heavy objects in the construction industry. In
recent years, numerous efforts have been made to
monitoring methods for crane-related activities in both
academia and industry. The first reason is safety, because
crane lifting is complex and risky on construction sites.
Cheng and Teizer (2011) found that many crane-related
safety incidents are associated with restricted visibility,
e.g., sight occlusion, poor weather, or lighting conditions.
Crane-related accidents usually cause casualties and
significant economic losses, so extra measures must be
taken. In addition, monitoring data could be visualized to
help operators reduce workload and improve work
efficiency, and the construction time is thus reduced.
Productivity assessment may be another reason for the
necessity of crane-related activities monitoring. Based on
the assessment result, further optimization work could be
done for the productivity improvement of the cranerelated construction process.

There are two common types of cranes, i.e., mobile cranes
and tower cranes. The most significant difference between
them is that one is mobile, and the other one is not. Due
to its fixed location and working range, monitoring or
sensing devices are constantly installed on the tower
crane. However, most existing monitoring methods are
challenging to implement on mobile cranes. Potential
reasons are the high mobility and flexibility of the mobile
crane. Therefore, how to achieve practical and convenient
monitoring for mobile cranes is still an unresolved issue.
With increasing focus on-site safety of the construction
industry and the emerging demand of digital construction
sites, developing a practical monitoring approach for
mobile cranes has become increasingly necessary and
urgent. In order to improve the situation, some
technologies, e.g., CCTV camera (Fang, et al., 2018)s and
IoT devices (Mijwil, et al., 2023), that are used in other
monitoring scenarios, are tried for monitoring or tracking
objects for the mobile crane during the lifting process.
However, the feasibility is not high for most methods. For
example, CCTV cameras, the most widely used tool to
monitor activities in construction sites, always lack depth
information, making it challenging to accurately capture
objects’ spatial position (Chen, et al., 2017). In addition,
quantization errors are widely reported to be the
shortcoming of conventional image-based video due to its
grid-based geometric representation. Sensitivity to light
conditions and the difficulty of multi-source data fusion
also make it unsuitable for monitoring mobile crane in
various complex environments. Second, IoT devices are
other type of tools for sensing objects on sites. IoT devices
are easy to locate, but various errors are widely reported.
Meanwhile, IoT devices cannot achieve dynamic detail
detection, such as rotation, which are significant for some
tracking purposes. Extra installation procedures for each
object will decrease productivity and increase risk, and
using many IoT devices would increase the cost. In
addition, the interaction of tracked objects and dynamic
environment cannot be easily reflected. Thirdly, the
Depth-RGB camera is also often reported to monitor or

track objects in sites potentially, but its short detection
distance restricts its application in outdoor scenarios.
Point clouds, a three-dimensional collection of data points
or coordinates, provide a new data form to produce highquality 3D reconstructions of the world (e.g.
reconstructing building information model in construction
site (Chen & Xue, 2023)). They provide more information
than two-dimensional pictures and are insensitive to light.
However, point clouds always refer to 3D point clouds
and the application of 3D point clouds is limited in
capturing some static objects due to only spatial
information contained in the data (Mirzaei, et al., 2022).
The advent of the 4D point cloud, consisting of three
dimensions and time information, has expanded the
application of the point cloud into a dynamic world (Silva,
et al., 2022). 4D point cloud (4DPC), a form of time-series
3D point clouds, is a rich data source demonstrating how
objects move against the background. 4DPC has unique
advantages against other motion tracking technologies,
e.g., various Internet of Things (IoT) and AI cameras
(Liang & Xue, 2022):
a) Real-time environment (including objects and
background) updating
b) Highly precise geometry updates
c) Higher adaptability in poor visibility
environments
d) Innate localization and mapping information
e) Innate capability of multi-/many-device data
fusion to eliminate visual blind spots
f) Lower
cost,
minimal
infrastructure
requirements, lower carbon footprints
g) Remote, non-destructive sensing
4DPC is a novel enabling data source that facilitates
precise 4D motion tracking. In contrast to conventional
camera/IoT, our methods have three characteristics:
i.
High-definition 4D motion data (cm-s-level
accuracy);
ii.
Precise depth information in a far range (up to
500m);
iii.
Simultaneous motion tracking of multiple
objects; and
iv.
Low cost of devices and easy operating.

2 Related works

Their primary purpose is to improve the efficiency and
safety of crane operations and reduce operator workload.
As listed in Table 1, this study reviews related to research,
and the purpose, contribution, research focus, target crane
type, methodology, and limitation are summarized. The
research generally has three types of goals, i.e. crane pose
estimation (Zhang, et al., 2012; Lee, et al., 2012; Zhong,
et al., 2014; Yang, et al., 2014; Roberts, et al., 2017), load
sway and rotation estimation (Fang, et al., 2018; Fang, et
al., 2016; Chian, et al., 2022), and object detection in
workspace (Li, et al., 2013; Chen, et al., 2017). Although
both crane monitoring and object detection are
investigated to help operators operate the crane in blind
areas in the study by Price et al. (2021), the two functions
are relatively separated and not well integrated into a
system.
Meanwhile, two types of sensing methods, i.e., sensorbased and vision-based, are used in these studies. For the
former, different sensors are serving for different sensing
goals. In specific, the UWB system, consisting of several
sensors and multiple tags, is used to estimate the crane
pose in the study by Zhang et al. (Zhang, et al., 2012), but
it is not practical due to various errors from different
sources and too many sensors required for full coverage
of activity range. Li et al. try to use GPS and RFID to
obtain positioning data of both site workers and the crane,
and using multiple tags and receivers decreases its
practicality. Zhong et al. (Zhong, et al., 2014) combined a
Wireless Sensor Network (WSN) and the Internet of
Things (IoT) to monitor the status of tower crane groups
to avoid collisions, but the interactions of tower cranes
with environment are not considered. Fang et al. (2016)
use a series of encoder and IMU sensors to monitor the
load object and visualize in the virtual platform. Except
for extensive use of sensors, another problem is the
dynamic change of the environment cannot be reflected
by a limited number of sensors. For the latter, CCTV
cameras or UAV camera are used to estimate crane pose
(Lee, et al., 2012; Yang, et al., 2014; Roberts, et al., 2017),
monitor load (Fang, et al., 2018) (Chian, et al., 2022), and
detect object (Chen, et al., 2017). The common limitation
stems from the data, such as not containing depth
information or sensitivity to lighting conditions and color.
In addition to shared drawbacks, UAV's endurance,
stability, and safety are also criticized.

2.1 Crane monitoring
In past research, numerous efforts have been put into
developing computer-aided crane monitoring systems.

Table 1 Summarization of related literature

Purpose and contribution

Research focus

Develop a UWB-based
system to track crane
boom movement, and
estimate crane pose near

crane pose
estimation

Sensing method Crane type
UWB

Mobile
crane

Limitation

Source

1) the installation space and device cost
of several sensors and multiple tags
would be a problem

(Zhang, et
al., 2012)

real-time for collision
avoidance

Develop a tower crane
navigation system to help
operators operate with
blind spots

2) the accuracy in an ideal environment
(open space) is 10cm, and it may be
worse in unordered sites
3) trajectory estimation is relatively
rough since it is based on linear
interpolation extrapolation of only two
points
crane pose
estimation

Video camera
&
sensors

Identify unauthorized
work or entrance of
personnel within a predefined risk zone

object detection in
the workspace

GPS
&
RFID

Develop a Safety
Management System to
monitor the status of
tower crane groups and
avoid collisions

crane pose
estimation

WSN
&
IoT

Understand construction
activity by tracking the
pose of the tower crane

crane pose
estimation

Video camera

1) too many kinds of sensors increase
the complexity of the proposed system
2) virtual environment (BIM model)
cannot BIM model cannot fully
represent the as-is site condition

(Lee, et al.,
2012)

Not limited 1) too many tags, receivers, and other
units are used;
to a
specific 2) installation complexity and high cost
would hinder its practical application;
type of
crane
3) signal strength is also a possible issue

(Li, et al.,
2013)

Tower
crane

Tower
crane

1) only the main body of the crane is
considered, but the site situation is
usually more complex

Tower
crane

1) large range view needs to be
covered, and the resolution of the object
will become lower, so the recognition (Yang, et al.,
accuracy will fluctuate due to many
2014)
factors, such as light condition
2) the recognition result is general

(Zhong, et
al., 2014)

Develop real-time proactive safety assistance
for mobile crane
lifting operations

load sway and
rotation estimation

Encoder sensors
&
IMU sensor

Mobile
crane

1) The pre-reconstructed site cannot
reflect the dynamically changing site
conditions
2) A large number of sensors are
required to sense the movement of the
crane in real time

Detect and Classify
Cranes for monitoring
crane-related safety
hazards

crane pose
estimation

UAV

Tower
crane

1) The endurance and stability of drones (Roberts, et
are obstacles to the practical application al., 2017)

Mobile
crane

1) positioning accuracy (0.1-0.4m) in
the ideal test environment is relatively
high
(Chen, et al.,
2017)
2) signal synchronization of camera and
LiDAR is complicated, and the error is
relatively large

Update real-time 3D crane object detection in
workspace
the workspace

Track crane load sway

Detect workers near the
crane load

TLS
&
Video camera

(Fang, et al.,
2016)

load sway and
rotation estimation

Not limited 1) only the 2D location of the load could
to a
be identified
(Fang, et al.,
specific
Video camera
2018)
2)
errors
highly
depend
on
the
quality
of
type of
the
image
crane

crane pose
estimation
&
load sway and
rotation estimation

1) sensor-part: positioning system has a
high reliance on data from noisy
encoders; large crane deflection
caused by the load leads to errors
(Price, et al.,
2) vision-part: positioning errors are
2021)
widely reported during irregular lighting
conditions and when the surrounding
environment contains objects that have a
similar color to the load

Sensors,
Camera,
IMU,
&
TLS

Tower
crane

Develop a novel method
to detect
and track the crane load
fall zone

load sway and
rotation estimation

Video camera

Tower
crane

1) the estimation accuracy heavily relies
on many factors, such as the quality of (Chian, et al.,
2022)
training data since it is based on deep
learning

#:UWB: Ultra Wideband; GPS: Global Positioning System; RFID: Radio-Frequency Identification; WSN: Wireless Sensor Network;

IoT: Internet of Things; IMU: Inertial Measurement Unit; UAV: Unmanned Aerial Vehicle; TLS: Terrestrial Laser Scanners.

In summary, existing sensor-based sensing methods in the
literature showed the following drawbacks:
1) Deploying and maintaining sensors in every
object onsite is complex and time-consuming,
which may affect productivity of the
construction process;
2) sensors are composed of many components
susceptible to damage, and low reliability may
hind its practice;
3) many high-precision sensors are expensive, and
the use of numerous sensors on-site may increase
the cost;
4) sensor technologies are sensitive to signal
interference and may not adapt to the complex
construction site with many existing physical
obstacles (e.g., existing buildings and
equipment);
5) it is difficult for sensors to capture the whole
environment with dynamical change, and hence
the interaction between tracking objects with the
existing site environment cannot be captured;
6) Synchronization and visualization of multisource data increase the practice complexity of
sensor technologies.
Meanwhile, current vision-based sensing methods have
other limitations:
1) Depth information is not contained in 2D image,
and hence vertical position information cannot
be precisely captured;
2) video errors caused by irregular lighting
conditions and similar colors make it unreliable
in daily use;
3) 3D laser scanning by TLS only captures the site
geometry at the time of scanning, and hence site
condition cannot be dynamically reflected in
real-time;
4) Multi-angle view fusion has not been well
resolved, which hinders its broad practice.
2.2 4DPC
Point clouds, a three-dimensional collection of data points
or coordinates, provide more information than twodimensional pictures and are insensitive to light (Bhople
et al., 2021). Conventional static point clouds have
already been widely used in many research and industrial
domains, such as surveying, electricity, construction, and
industry, due to their excellent ability to represent our
three-dimensional world. In the construction industry, the

3D point cloud is currently used for various aims such as
as-built building reconstruction and digital twin city (Wu,
et al., 2021; Xue, et al., 2019; 2020). It should be noted
that the applications of 3D point clouds are limited in
capturing non-moving or assumingly non-moving
objects.
As a window into our ever-evolving 3D environment,
4DPC are widely used in robotics and augmented reality
systems. Point cloud sequences play an important role in
understanding environmental changes and supporting
interactions with the world that are difficult to describe
with 2D images or static 3D point clouds due to their
ability to record movements in physical space. In order to
more accurately simulate the world, respond to changes in
the environment, and interact with it, an intelligent agent
must handle this kind of data with great precision.
4DPC has gained popularity for some reasons. First, the
ability to comprehend a changing 3D world is essential for
robotic agents and various other applications. 4DPC has
enabled many innovative studies, such as how a plant
grows (Li, et al., 2013) and high-definition human
motions (Fan, et al., 2021), as shown in Figure 1. In
addition, various identification tasks, such as calculating
a moving object's acceleration or identifying human
activities, benefit from temporal data sequences longer
than two frames (Fan, et al., 2021). 4D point cloud has
also been widely utilized in robotic SLAM, autonomous
driving, and video-assisted training of athletes.

Figure 1 Example 4DPC-enabled studies. (a) how plant grows
(Li, et al., 2013)—permission requesting, (b) high-definition
human motions (Fan, et al., 2021)—permission requesting

In contrast to grid-based RGB video, where points
regularly emerge over time, point cloud video displays
irregularities and lacks order along the spatial dimension.
Therefore, numerous efforts have been engaged in
processing the 4DPC data. Two existing methods, i.e.,

voxelization-based (Choy, et al., 2019) and point-setbased, explored by many researchers to process 4DPC
data.

3 Case study
3.1 Case description
A pilot study was conducted on an infrastructure project
in Hong Kong, as shown in Figure 4. The case project was
the 3rd Sassoon Road footbridge project that links a new
Campus building to the student residents across Sassoon
Road. Two major steel tie-beams were hoisted and
installed in the mid-night of 28 June 2022, and Figure 2
shows the site condition before hoisting, the hoisting
process of component 1, the completed status of the
component 1 hoisting, and completed status of two
components hoisting, respectively. This project uses a
mobile crane since it is a temporary infrastructure project

and the road is only temporarily closed for construction at
night.
The case project had characteristics and requirements in
hoist efficiency, collision risk, and personnel safety. First,
the process needs to be completed swiftly before dawn to
resume day traffic from the temporary close of the road.
As the most important and challenging work in this
project, hoisting efficiency will directly affect the
construction time. In addition, there were already building
works on both sides of the road, and the facades are
expensive glass curtain walls. A collision during the
hoisting process will cause economic losses. Therefore,
effective measures should be taken to reduce the potential
collision risk. Furthermore, the workers working on other
processes on-site may have entered the risky zone where
the hoisted objects may fall due to negligence of steel
beam at high height during the hoisting process.

Figure 2 construction site condition of case study

3.2 4DPC devices and setup
To address the limitations mentioned in Sec. 2.1, this
study proposes and validates a scheme that 4DPC
technology is used to achieve the goal of monitoring all
crane-related activities in real-time, including estimating
crane pose, tracking load, and detecting objects on sites.
To achieve that, a novel 4DPC sensing equipment for
collecting high-definition 4D motion data from
construction environments is developed in this study. The
target spatial accuracy is cm-level, while the temporal
accuracy is 0.5s-level. As shown in Figure 3(b), the device
has four essential modules, i.e., (Ⅰ) Livox Mid-70 sensor,
(Ⅱ) controller (Raspberry Pi 4), (Ⅲ) LED monitor, (Ⅳ)
USB drive. The Livox Mid-70 has 70.4° circular fov, 5
cm minimum detection range, and 2 cm range precision.
Raspberry Pi 4 controller transmits 4DPC data to remote
server via WiFi/4G/USB. The proof-to-concept test of the
device is conducted in the pilot study. In order to cover
the whole site, two LiDAR devices in Figure 3 (b) were
installed on different locations, i.e., one is in the ground
floor and the other is in the 4th floor (shown in Figure
3(a)), to collect 6 hours of 4DPC data (format: lvx; size:
80MB/min).

Figure 3 (a) Installation illustration of 4DPC devices on site;
(b) Components of our 4DPC device: (Ⅰ) Livox Mid sensor, (Ⅱ)
controller (Raspberry Pi 4), (Ⅲ) LED monitor, (Ⅳ) USB drive

3.3 Methods
The most popular processing methods of 4DPC in existing
research are based on deep learning. The impetus for this
trend is mainly based on the large amount of data that can
be collected. In this study, we only collected 1
construction scenario; hence, deep learning is not
applicable. Therefore, a rule-based object detection
method on 4DPC data is proposed to monitor target

components in real time effectively. As shown in Figure
4, a cyclic processing workflow is determined:
1) 4DPC data input: Each frame of data is input in
chronological order;
2) Background removal: using random sample
consensus (RANSAC) algorithm (Schnabel, et
al., 2007) is used to detect and delete plane of
existing ground and walls
3) Clustering:
using DBSCAN clustering
algorithm (Ester, et al., 1996) to cluster the
remaining point cloud.
4) First round match: judging whether the point
cloud clustering results match the objects with
known geometric dimensions. If all known
components match, go to step 7. If not, continue
to step 5.
5) crop point cloud: using bounding box of
previous frame to crop the whole point cloud

(based on the assumption that there will be no
large displacement of the member in a very short
time-0.5s)
6) Second round match: judging whether the crop
result of point cloud match the objects with
known geometric dimensions. If known
components match, go to step 7. If not, the result
of the previous frame will be assigned to the
current frame (based on the assumption that
there will be no large displacement of the
member in a very short time-0.5s).
7) Result saving: Saving related point cloud
segmentation
results
and
creating
a
corresponding bounding
box for
the
corresponding component
The program ends when all the time series data are
processed.

Figure 4 logic flow of rule-based detection method for time series 4DPC data

3.4 Preliminary results and discussion
As shown in Figure 5, the two steel beam components can
be precisely detected in more than 95% of the time using
the rule-based method described in Sect 3.3. The 4DPC
data processing time per minute is within 1s.
Figure 5(a), (b), (c), and (d) illustrate the motion detection
result of two beams from two views at different times, and

different typical statuses of beams, i.e., remaining on the
transport vehicle, being placed on the ground, hoisting in
the air, are all covered, or completing hoisting. However,
there are also some time frames that beams cannot be
tracked. Detailed analysis suggests the ineffective capture
is that the point cloud data is too sparse or no data points
exist due to small sensing area of components at specific

view angles, long sensing distances, or some physical
obstacles. These problems can be attributed to the limited
number of 4DPC devices used in this pilot study and the
low point cloud density provided by Livox. Increasing the
number of LiDAR devices and using advanced LiDAR
devices providing higher point cloud density would
improve the robustness of 4DPC data. Obtaining the
spatial-temporal information of beams will help further
achieve a set of proactive motion prediction and
optimization applications. In addition, it could facilitate
the productivity assessment of mobile cranes, and future
improvement may be based on these data records.
3.4 Limitation
While this study's proposed method effectively captures
targeted objects, its rule-based approach is limited to the

specific scenario in which it was tested. Therefore, future
research is necessary to explore its actual deployment
and extended application in different conditions.
Furthermore, the rule-based method can only detect
known objects since rules must be set with prior
knowledge of components. Also, due to the angle of the
equipment installation, the object cannot be fully
scanned, which limits the reflection of the true detailed
geometry of the component. Detailed modeling requires
additional effort. Finally, it should be noted that the
4DPC data collected from two devices was registered
manually, which is a time-consuming process and
hinders its application in mobile situations..

Figure 5 typical detection result of 4DPC data

4 Conclusion and future work
This paper evaluated the existing studies focusing on
sensing or monitoring crane-related activities, including
estimating crane pose, tracking load, and detecting objects
in workspace. Meanwhile, the limitation of sensor-based
and vision-based sensing methods are analyzed and
summarized. In order to overcome shortcoming of
existing methods, a state-of-art 4DPC sensing and
monitoring method is proposed. A rule-based object
detection method is developed with two prototype devices
of 4DPC sensing, and a proof-to-concept test was
conducted on a footbridge construction project. In the
pilot study, two 4DPC devices covered the whole
construction site for 6 hours. The promising preliminary
results suggested that target beam components were
precisely captured in different statuses. The spatiotemporal data series obtained from the tracking results
could generally satisfy the goal of monitoring load.

There are three directions for future works. One is to
develop more efficient and robust methods, such as some
ML-based methods (Zhang, et al., 2019; Liang & Xue,
2023), to match or detect objects from the collected data.
Another is to explore a set of proactive motions prediction
and optimization applications. The last is to integrate
obtained spatial-temporal information of all objects and
workers to achieve dynamic site management.

Acknowledgement
The work presented in this paper was supported by The
Shenzhen Science and Technology Innovation
Commission (SZSTI) (No. SGDX20201103093600002).

Author Contributions
Conceptualization, D.L. and F.X.; Methodology, D.L.;
Software, W.Y.; Validation, Z.C., L.K. and S.C.; Formal
Analysis, D.L.; Data Curation, F.X.; Writing – Original

Draft Preparation, D.L.; Writing – Review & Editing,
F.X.; Visualization, D.L. and F.X.; Supervision, F.X.;
Funding Acquisition, F.X.

References
Cheng, T. & Teizer, J., 2011. Crane operator visibility of
ground operations. Seoul, Korea, s.n., pp. 699-705.
Chen, J., Fang, Y. & Cho, Y. K., 2017. Real-time 3D
crane workspace update using a hybrid visualization
approach. Journal of Computing in Civil Engineering,
31(5), p. 04017049.
Chen, S.-H. & Xue, F., 2023. Automatic BIM detailing
using deep features of 3D views. Automation in
Construction, Volume 148, p. 104780.
Chian, E. Y. T., Goh, Y. M., Tian, J. & Guo, B. H., 2022.
Dynamic identification of crane load fall zone: A
computer vision approach. Safety Science, Volume
156, p. 105904.
Choy, C., Gwak, J. & Savarese, S., 2019. 4D spatiotemporal convnets: Minkowski convolutional neural
networks. s.l., s.n., pp. 3075-3084.
Ester, Kriegel, H.-P., Sander, J. & Xu, X., 1996. A
density-based algorithm for discovering clusters in
large spatial databases with noise. KDD, 96(34), pp.
226-231.
Fang, Y. et al., 2018. Vision-based load sway monitoring
to improve crane safety in blind lifts. Journal of
Structural Integrity and Maintenance, 3(4), pp. 233242.
Fang, Y., Cho, Y. K. & Chen, J., 2016. A framework for
real-time pro-active safety assistance for mobile crane
lifting operations. Automation in Construction,
Volume 72, pp. 367-379.
Fan, H., Yang, Y. & Kankanhalli, M., 2021. Point 4d
transformer networks for spatio-temporal modeling in
point cloud videos. s.l., s.n., pp. 14204-14213.
Lee, G. et al., 2012. A BIM- and sensor-based tower crane
navigation system for blind lifts. Automation in
Construction, Volume 26, pp. 1-10.
Liang, D. & Xue, F., 2022. Applications of 4D Point
Clouds (4DPC) in Digital Twin Construction: A
SWOT Analysis. s.l., s.n., p. (in press).
Liang, D. & Xue, F., 2023. Integrating automated
machine learning and interpretability analysis in
architecture, engineering and construction industry: A
case of identifying failure modes of reinforced
concrete shear walls. Computers in Industry, Volume
147, p. 103883.
Li, H., Chan, G. & Skitmore, M., 2013. Integrating real
time positioning systems to improve blind lifting and
loading crane operations. Construction Management
and Economics, 31(6), pp. 596-605.
Li, Y. et al., 2013. Analyzing growing plants from 4D
point cloud data. ACM Transactions on Graphics,
32(6), pp. 1-10.

Mijwil, M. M., Hiran, K. K., Doshi, R. & Unogwu, O. J.,
2023. Advancing Construction with IoT and RFID
Technology in Civil Engineering: A Technology
Review. Al-Salam Journal for Engineering and
Technology, 2(2), pp. 54-62.
Mirzaei, K. et al., 2022. 3D point cloud data processing
with machine learning for construction and
infrastructure applications: A comprehensive review.
Advanced Engineering Informatics, Volume 51, p.
101501.
Price, L. C., Chen, J., Park, J. & Cho, Y. K., 2021.
Multisensor-driven real-time crane monitoring system
for blind lift operations: Lessons learned from a case
study. Automation in Construction, Volume 124, p.
103552.
Roberts, D., Bretl, T. & Golparvar-Fard, M., 2017.
Detecting and classifying cranes using cameraequipped UAVs for monitoring crane-related safety
hazards. Seattle, Washington, s.n., pp. 442-449.
Schnabel, R., Wah, R. & Klein, R., 2007. Efficient
RANSAC for Point-Cloud Shape Detection.
Computer graphics forum, 26(2), pp. 214-226.
Silva, M. F. et al., 2022. 3D structural vibration
identification from dynamic point clouds. Mechanical
Systems and Signal Processing, Volume 166, p.
108352.
Wu, Y., Shang, J. & Xue, F., 2021. RegARD: Symmetrybased coarse registration of smartphone’s colorful
point clouds with CAD drawings for low-cost Digital
Twin Buildings. remote sensing, 13(10), p. 1882.
Xue, F., Lu, W., Chen, K. & Webster, C. J., 2019. BIM
reconstruction from 3D point clouds: A semantic
registration approach based on multimodal
optimization and architectural design knowledge.
Advanced engineering informatics, Volume 42, p.
100965.
Xue, F., Lu, W., Chen, Z. & Webster, C. J., 2020. From
LiDAR point cloud towards digital twin city:
Clustering city objects based on Gestalt principles.
ISPRS Journal of Photogrammetry and Remote
Sensing, Volume 167, pp. 418-431.
Yang, J., Vela, P., Teizer, J. & Shi, Z., 2014. VisionBased Tower Crane Tracking for Understanding
Construction Activity. Journal of Computing in Civil
Engineering, 28(1).
Zhang, C., Hammad, A. & Rodriguez, S., 2012. Crane
pose estimation using UWB real-time location system.
Journal of Computing in Civil Engineering, 26(5), pp.
625-637.
Zhang, J., Zhao, X., Chen, Z. & Lu, Z., 2019. A review of
deep learning-based semantic segmentation for point
cloud.
IEEE
Access,
Issue
10.1109/access.2019.2958671 , pp. 179118-179133.
Zhong, D., Lv, H., Han, J. & Wei, Q., 2014. A Practical
Application Combining Wireless Sensor Networks
and Internet of Things: Safety Management System

for Tower Crane Groups. sensors, Volume 14, pp.
13794-13814.