Easy in the Cloud! How UAV Photogrammetry Point Cloud Generation is Changing Infrastructure and Building Surveying

Table of Contents

• What does it mean to generate point clouds with UAVs (drones)?

• Advantages of drone (UAV) point cloud surveying

• Benefits of generating point clouds in the cloud

• Steps to start UAV point cloud generation in the cloud

• Key points for successful UAV point cloud generation

• Main use cases of point cloud data

• Recommendation: simple surveying with LRTK

• FAQ

In recent years, DX (digital transformation) has rapidly advanced in the construction and surveying industries, and with the development of drones and 3D technologies, the term "point cloud data" is increasingly heard on job sites. In infrastructure (civil engineering) and building surveying sites as well, there is a growing movement to obtain terrain, infrastructure structures, and buildings as three-dimensional point cloud data from drone photographs, and to utilize them for construction measurement and as-built management. However, beginners who want to start surveying with drones may feel anxious, wondering "Where should I start?" or "Do I need expensive equipment and complex software?" In fact, recently, by leveraging cloud services, it has become easy to start generating point clouds from drone photos even without specialized knowledge or a high-performance PC. The fact that high-precision point cloud models can now be obtained easily in the cloud is truly a new norm for surveying in the DX era. In this article, we explain in an easy-to-understand manner, from the basics of point cloud data generation using drones (UAVs), to the advantages of using the cloud, concrete startup procedures, and tips for success. At the end of the article, we also introduce the notable tool "LRTK" that allows even beginners to easily experience 3D surveying. If you are interested in drone-based point cloud measurement but have not taken the first step, please use this as a reference.

What does it mean to generate point clouds with UAVs (drones)?

First, point cloud data refers to three-dimensional data that represents real-world objects as a collection of countless points. Each point includes X, Y, and Z position coordinates, and when the set of points is visualized on a PC, the shape of the object can be reproduced three-dimensionally.

So how can point cloud data be generated by drones (UAVs)? The common method is called photogrammetry. By taking many photographs of terrain and buildings from various angles from the air using a camera mounted on a drone and processing these photos with dedicated analysis software or a cloud service, it is possible to obtain high-density 3D point cloud data. The technique matches common feature points appearing in multiple images and calculates the three-dimensional positions of points based on the parallax between photos taken from different viewpoints.

There is also a method of mounting a lightweight laser scanner (LiDAR) on a drone to directly acquire point clouds by laser ranging. Laser measurement has advantages such as being able to measure ground surfaces beneath vegetation, but equipment costs are very high and operation is difficult, making it a higher barrier for beginners to introduce immediately. In contrast, photogrammetry can be undertaken with a commercially available camera-equipped drone and analysis software (or a cloud service), making it an ideal approach for first-time point cloud generation.

Advantages of drone (UAV) point cloud surveying

Using drones for point cloud surveying offers many advantages compared with traditional methods. Here are four main benefits.

• Efficiency and labor savings: With aerial photogrammetry using drones, large-area surveys that used to require multiple people and several days can be completed in a short time by a single person. Because detailed data for a wide site can be acquired in a single flight, “missed measurements” are reduced and additional surveying work is minimized. In the construction industry, where chronic labor shortages are a challenge, the ability to greatly streamline operations and reduce personnel is highly attractive.

• High-accuracy, comprehensive measurement: Point cloud data records objects with countless points, so it provides far more precise and comprehensive information than methods that measure only limited points. Since even millimeter-level fine shapes can be preserved as digital data, remeasuring arbitrary locations later is easy. Even if no drawings remain on site, accurate 3D models can be produced, aiding quality control and improving design and construction accuracy.

• Improved safety: Drone photogrammetry and laser scanning are remote, non-contact measurements, allowing safe data acquisition even in hazardous areas that people cannot enter. High-altitude or steep-slope surveys can be performed remotely with drones, significantly reducing the risk of accidents during operations. Tasks that previously required personnel working at height or road traffic control can often be minimized, greatly contributing to on-site safety management.

• Smoother consensus building through 3D data sharing: Point clouds captured by drones can be shared with stakeholders in the cloud as 3D models. Even supervisors or clients located remotely can view the site in three dimensions via PCs or tablets, facilitating smoother communication. Information that was difficult to convey with flat drawings or photos can be understood at a glance with 3D models. During as-built inspections or design change meetings, explanations using 3D data are more persuasive and can lead to quicker consensus.

Because of these benefits, the Ministry of Land, Infrastructure, Transport and Tourism is promoting the use of ICT technologies such as drone surveying under the "i-Construction" initiative, and point cloud technology is becoming a symbolic presence in construction industry DX. The norms of surveying in the infrastructure and building sectors are truly beginning to change.

Benefits of generating point clouds in the cloud

Next, let’s look at what advantages there are to performing point cloud generation from drone photos on cloud services. By leveraging the cloud, beginners without specialized knowledge can more easily obtain high-quality results.

• No need for a high-performance PC: Generating point cloud data normally requires a PC capable of processing many high-resolution images, but with cloud services, heavy computations are executed entirely on the server side. On-site, operations can be performed with a typical laptop or tablet, eliminating the need to prepare an expensive workstation yourself.

• Easy to set up and operate: There is no need to install dedicated software or perform complex configuration—just upload the captured images via a web browser and click a button to start analysis. Simple UIs that even beginners can use without hesitation are provided, greatly reducing the time needed to learn software operation from scratch.

• Always access to the latest technology: Because analysis algorithms and features are updated daily on the cloud service side, users always enjoy the latest versions of the technology. There is no need to update software yourself or worry about version compatibility. New point cloud processing technologies and AI-based automatic classification functions will be available on the cloud as soon as they are added.

• Easy for teams to use: With data on the cloud, results of point cloud generation can be easily shared and used by multiple people. By inviting stakeholders to the same project, each person can view and measure 3D data via the browser without having to prepare software individually. Remote team members can share the situation in real time, enabling smooth collaborative work across distant sites.

• Cost advantages: Cloud services are also attractive because they can be introduced with low initial investment. Previously, purchasing surveying software or hardware could cost hundreds of thousands of yen, but with the cloud you can use what you need, when you need it. Many services adopt monthly subscription or pay-as-you-go models, making cost management by project straightforward. For small-scale operators who do not use the service frequently, the cloud offers an economical option to use the latest technology without waste.

Steps to start UAV point cloud generation in the cloud

Now, let’s follow the basic steps for generating point clouds from drone photos using the cloud. The general flow is as follows.

• ① Equipment preparation and flight planning: First, prepare the drone to be used (preferably equipped with a high-resolution camera) and necessary equipment such as the transmitter and batteries. Plan flight routes so they efficiently cover the entire area you want to survey. Set the photo overlap rate to 70% or more in front-back and left-right directions to obtain high-quality point clouds. Also check the weather and wind speed in advance and choose a time when it is safe to fly. Be sure to confirm drone flight regulations such as whether the area is a no-fly zone and whether permits are required.

• ② Photographing with the drone: Fly the drone autonomously or manually according to the plan and take photos of the target area. The key is to maintain the set altitude and camera angle and photograph the ground thoroughly. Generally, you fly in parallel routes in a zigzag pattern while continuously shooting straight down. When there are buildings or structures, combine oblique shots from angled directions to reduce blind spots and obtain a more complete point cloud model. During the flight, always monitor battery levels and aircraft condition, and prioritize safety without overextending operations.

• ③ Uploading photo data to the cloud: After shooting, access the cloud point cloud processing platform and create a new project. Upload the many photos taken on site to the service. If there are many images, a large amount of data communication will occur at once, so it is advisable to use a stable, high-speed Internet connection if possible. If you used a GPS-equipped drone, each image file contains geotags, so the cloud will automatically perform initial image alignment. If you have known survey control points (GCPs), upload them at this stage to improve accuracy during post-processing.

• ④ Point cloud generation on the cloud (processing): After uploading the photos, start point cloud generation analysis on the cloud. Follow the service prompts to confirm settings, then click the start analysis button. Image analysis (feature matching and 3D reconstruction) runs automatically on the server, and even projects with dozens to hundreds of photos can produce point cloud data in a relatively short time thanks to the cloud’s high-performance computing environment. Processing time can range from several tens of minutes to several hours depending on data volume, but progress can be checked in the browser. Users can continue other work while waiting for processing to finish, which is efficient in terms of productivity.

• ⑤ Review and utilization of results: Once processing is complete, review the generated point cloud data in the cloud. Use the platform’s 3D viewer to freely rotate and zoom the point cloud model to check quality. Carefully confirm whether the terrain and structures are correctly reproduced and whether there are any missing parts. Some services offer filtering functions to remove unnatural noise points. If there are no problems, download the point cloud data as needed and import it into CAD or surveying software for use. Some cloud services also allow on-the-fly volume calculations or cross-section generation, and generated orthophotos (composite top-down images) and 3D mesh models are useful deliverables on site. Share the completed deliverables with stakeholders on the cloud or attach them to reports to support infrastructure and building site operations.

Key points for successful UAV point cloud generation

When attempting point cloud generation from drone photos for the first time, there are several points to keep in mind. Paying attention to the following will help you obtain higher-quality 3D models.

• Ensure sufficient photo overlap: The accuracy and density of point cloud generation are greatly affected by the overlap between captured photos. Ensure a front-back and left-right overlap rate of 70–80% or more so that the same ground area is captured clearly in adjacent photos. Insufficient overlap can cause gaps in the point cloud or reduced accuracy.

• Consider weather and shooting conditions: In photogrammetry, it is important to capture bright, sharp images as much as possible. Choose times with sunlight and avoid flights in rain or strong wind. Also avoid leaving camera settings entirely on automatic—fix focus and use a fast shutter to prevent blur. Shooting in manual mode to keep exposure consistent across the series of photos is also effective to avoid extreme exposure differences.

• Implement higher-accuracy measures when necessary: In cases where elevation and coordinate accuracy are particularly important, consider improving positioning accuracy using RTK-equipped drones or ground control points (GCPs) installed on the ground. Using these allows you to attach precise latitude, longitude, and elevation information to the generated point cloud and can achieve accuracy comparable to official surveying results. Although initial cost and effort increase, it is worthwhile for public surveys and infrastructure design tasks that require high precision.

• Practice on small areas first: Rather than attempting a large-scale site immediately, start with a small area or familiar structure. By experiencing the full workflow with a small dataset and verifying the results, you will discover points for improvement and tips for the next time. Gradually expanding the target area and project scale reduces the risk of failure while steadily improving skills.

• Comply with regulations and ensure safety management: There are various legal rules for drone flights. Always check in advance for no-fly zones, required permit applications, and aircraft registration (in Japan, aircraft weighing 100 g or more must be registered). Also thoroughly implement safety procedures such as battery charge management, GPS status checks, and measures to prevent third parties from entering the site. On site, pay close attention to the surroundings and operate cautiously—carrying out operations with zero accidents is the most important priority.

Main use cases of point cloud data

High-precision point cloud data generated by drones is useful in many situations in the infrastructure and building fields. Below are common use cases on site.

• Earthwork volume calculation and as-built management: From acquired point cloud data, you can calculate volumes of embankment and excavation and accurately record completed terrain as as-built drawings. For example, by scanning the ground before and after construction with a drone and comparing, you can accurately determine the amount of soil removed or brought in. You can also extract arbitrary cross-sections from point clouds to verify whether constructed elements match design dimensions and shapes. Point cloud utilization has made high-precision as-built management that was previously difficult much easier.

• As-built mapping and design review: Drone point clouds serve as base data for high-density as-built surveys. From the acquired 3D point cloud, you can create plan views and longitudinal/cross-sectional drawings as needed, helping designers accurately grasp on-site conditions. It is also possible to overlay design models (CAD or BIM data) on point clouds to check for clashes or to simulate the validity of planned routes. Intuitive 3D data helps identify issues that may be overlooked in traditional 2D drawings.

• Construction progress management and stakeholder sharing: Regularly recording construction progress with drone point clouds allows you to track site changes over time. By overlaying point clouds from different times, you can immediately see where and how much changes have occurred, which is useful for schedule management and progress reporting. Sharing these data via the cloud with clients and team members also delivers a vivid, up-to-date sense of site conditions to stakeholders who cannot visit. Presenting 3D models in regular meetings helps speed up consensus building.

• Infrastructure inspection and disaster response: Point cloud data obtained by drone-mounted cameras or LiDAR is used for maintenance of infrastructure such as bridges, dams, and tunnels, and for rapid situational assessment in disaster areas. Because data can be collected safely in places where people cannot enter, it is powerful for monitoring steep slopes at risk of collapse or measuring terrain changes after flooding. Skills learned from drone photogrammetry in peacetime will be useful in such special cases when urgent response is required.

Recommendation: simple surveying with LRTK

LRTK is an innovative solution developed to allow beginners to easily start 3D surveying. Based on the idea of "smartphone + RTK," by attaching a small positioning device called the LRTK Phone to a handheld iPhone and launching the dedicated app, high-precision point cloud measurement becomes possible. It realizes centimeter-class high-precision positioning (real-time kinematic: RTK) that previously required special equipment on a smartphone, allowing accurate position information to be attached to photos and LiDAR scans taken with the smartphone. The convenience of being able to perform site surveying alone with a single small device that fits in a pocket truly merits the term "simple surveying." Even those without surveying experience can obtain 3D point cloud data simply by following the app’s guidance and pointing the smartphone, greatly lowering the barrier to on-site introduction.

Furthermore, LRTK integrates with cloud services to enable seamless use of acquired data. For example, if you upload point cloud data obtained with LRTK or numerous photos taken by a drone to the LRTK cloud, a point cloud model with positional coordinates is automatically generated and can be edited and shared in the browser. Because you can handle large-scale 3D data without a high-spec PC at hand, LRTK supports everything from small sites to large-scale projects. LRTK, which lets you try the latest 3D technology while keeping initial costs down, is an ideal tool for promoting on-site DX with the approach of "start small and expand gradually." If you have been hesitant about using point cloud data, especially as a beginner, consider taking a simple first step by using LRTK.

(Note: The phrase "cm精度" is used here to indicate cm level accuracy (half-inch accuracy): cm level accuracy (half-inch accuracy).)

FAQ

Q: What equipment and preparations are needed to start drone surveying? A: Basically, you need a drone equipped with a high-resolution camera and accessories such as a transmitter (controller) and batteries to fly the drone safely. Also prepare to register for a cloud service (or dedicated photogrammetry software) to process the captured photos into point clouds. If you need to align survey results to map coordinate systems, using an RTK-capable drone or ground control points (GCPs) will improve positional accuracy. Additionally, check in advance whether the planned flight area falls under restricted airspace and, if necessary, complete legal procedures such as applying for permits to the Ministry of Land, Infrastructure, Transport and Tourism and registering the aircraft (in Japan, aircraft weighing 100 g or more must be registered).

Q: Can beginners generate point cloud data with drones? A: Yes. In recent years, drone autopilot apps and cloud service interfaces have become sophisticated, and beginners can obtain point cloud data by following basic procedures. However, safe operation and legal compliance are required for flying aircraft, so we recommend practicing and gaining experience on relatively simple projects at first. Even if things do not go as planned initially, repeated trials on small sites will allow you to master the techniques and improve gradually.

Q: Are there disadvantages to using cloud services? A: The main considerations are dependence on Internet connectivity. Because you need to upload many high-resolution photos, an unstable connection can slow processing and increase communication costs. Also, cloud service fees are ongoing. However, many services use pay-as-you-go models with no initial cost, so they can be cheaper overall than purchasing expensive on-premises software. If data confidentiality is a concern, choose a reputable service with a strong track record and check the terms of use regarding uploaded data handling.

Q: What level of accuracy can be achieved with point cloud data? A: Accuracy varies depending on shooting methods and equipment, but with photogrammetry—if overlap is sufficient and sharp images are obtained—you can typically expect horizontal errors on the order of several centimeters to a dozen or so centimeters, and vertical errors around 10-20 cm (3.9-7.9 in). By using RTK-capable drones or GCPs and applying appropriate corrections, accuracy within a few centimeters is achievable. Note that accuracy decreases in areas with dense vegetation or on water surfaces where feature points are scarce. Depending on required accuracy, consider using RTK or GCPs, and in some cases combining with LiDAR.

Q: Are qualifications or permits required to operate drones? A: In Japan, since 2022, aircraft weighing 100 g or more must be registered. Additionally, certain operations—such as beyond-visual-line-of-sight (BVLOS) flights, nighttime flights, or flying over densely inhabited districts (DID)—require flight permission/approval from the Ministry of Land, Infrastructure, Transport and Tourism. A national qualification (license) system for unmanned aircraft pilots has also started for some flight categories. However, basic flights within visual line of sight may still be conducted without a license in many cases. Laws and regulations may change, so always check the latest guidelines from the Ministry of Land, Infrastructure, Transport and Tourism before flying and follow the appropriate procedures.