Automatically Generate Point Cloud Data Just by Uploading Photos | How to Choose and Implement Cloud Processing Services

Table of Contents

• What is a cloud service that can automatically generate point cloud data from photos

• Conventional point cloud generation methods and their challenges

• New point cloud generation enabled by cloud processing services

• Benefits of leveraging cloud services

• How to choose a cloud service for point cloud generation

• Points for implementation (tips for success and precautions)

• Recommendation for simple surveying using LRTK

• FAQ

In recent years, the use of 3D point cloud data has been attracting attention in construction and surveying. By digitally reproducing terrain and structures as countless collections of points (point clouds), it is possible to handle detailed information that conventional planar drawings and photographs cannot provide. However, acquiring point cloud data requires specialized laser scanners and expensive software, and has long been considered a high hurdle for beginners.

What has emerged is a cloud service that can automatically generate point cloud data simply by uploading photos. By simply uploading photos taken with a smartphone or a drone to the cloud, powerful backend processing servers analyze the images and automatically produce a 3D point cloud model. In this article, we explain in detail how these cloud point-cloud processing services work, how they differ from traditional methods, the benefits of adopting them, points to consider when choosing a service, and tips for implementation. Finally, we will also introduce the popular tool LRTK, which allows even beginners to easily start 3D surveying.

What are cloud services that can automatically generate point cloud data from photos

First, let's go over the technology for generating point cloud data from photos. The technique of reconstructing an object's shape from multiple photographic images is called photogrammetry (photogrammetry), and recent advances in computer vision technology have dramatically improved both accuracy and speed. If you capture a large number of photos that cover the subject—such as images of the ground taken from above by a drone or photos of a building shot from various angles with a smartphone—those images can be analyzed to find common points and reconstruct a 3D model in point cloud format.

Cloud-based point cloud services carry out this photogrammetry process on servers over the internet. Users simply access the service from a web browser and upload the many photos they have taken. Advanced algorithms then run automatically on the cloud side and generate 3D point cloud data from the uploaded set of photos. For example, if you send hundreds of aerial photos taken by a drone to a dedicated site, a detailed point cloud model of the entire site will be created in the cloud after a few hours. The appeal is the ease of generating high-precision 3D data from anywhere with just an internet connection, without having to install software on your own computer or run heavy computations.

Traditional point cloud generation methods and their challenges

So, how were point cloud data created when not using the cloud? Traditionally, two main methods were common. One was to bring a ground-based laser scanner (LiDAR) to the site and perform measurements. The other was to analyze photos taken by drones or single-lens cameras with photogrammetry software to generate point clouds. However, these methods presented a high barrier to entry for beginners.

First, the former laser scanner measurements require equipment that is extremely expensive and specialized. While they can produce precise 3D point clouds, purchasing the hardware and receiving operational training incurs significant costs. The latter method using photogrammetry software may at first glance seem easy and doable with a consumer camera, but in reality it required a high-performance PC and advanced software. To process the hundreds of high-resolution photos, a workstation with vast memory and GPU power is indispensable. The software itself is also costly due to its specialized purpose, and its interface has many configuration options, making it difficult for amateurs to use.

Furthermore, with traditional methods, time and effort were also significant challenges. For example, after taking photos on site, it was not uncommon to return to the office and leave a computer running the analysis overnight, with the point cloud model only finally completed the next day. If parameter adjustments were required during the analysis, a specialist operator had to be on hand each time. Even after completing the on-site shooting, there was a time lag before the final data could be obtained.

Also, the handling of the generated point cloud data brought its own difficulties. Point cloud files are very large (sometimes several GB or more), and simply storing and sharing them is a challenge. Email attachments are, of course, difficult, and there were occasions when we had no choice but to rely on primitive methods, such as copying them to an external hard disk and handing it over. The fact that it was difficult to easily share and view the 3D data we had obtained with all stakeholders was a major bottleneck.

New Point Cloud Generation Expanding with Cloud Processing Services

To address the challenges described above, services that can complete point cloud generation entirely in the cloud have emerged. With cloud-based point cloud services, users only need to upload photos, and all the heavy processing is automatically handled server-side. Even without a high-performance PC on hand, you can leverage the cloud's powerful computing resources over the internet, enabling large-scale photo analysis to be completed in a short time.

Using the cloud service is very simple. First, register an account on the web and create a new project. Then upload multiple photo files taken in the target area all at once. Once the upload is complete, simply click the button to start the analysis. Behind the scenes, photogrammetry algorithms (feature point matching and 3D reconstruction processing) run in parallel, and a point cloud model is generated from the uploaded image set in one go. When processing finishes, the project on the cloud stores the point cloud data as a finished product, and users can view it in a web browser. Some services also allow you to download the data as needed or to measure and edit it directly in the browser.

Thus, by leveraging cloud processing services, point cloud generation—which until now was difficult without specialized environments—becomes accessible to anyone. Because it can be used without depending on a specific computer setup, you can obtain the same 3D models from a field laptop or an office desktop. It is also possible to check results from a tablet while out of the office. If you upload photos to the cloud during spare moments, the analysis can be completed during lunch breaks or travel time, enabling a more efficient workflow. By turning the point cloud generation process itself into a service, it is helping accelerate on-site DX (digital transformation).

Benefits of Using Cloud Services

Generating point clouds in the cloud offers many advantages over traditional methods. Here, let's outline the main benefits.

• Time savings and labor reduction: High-speed processing by powerful servers greatly reduces the time required for point cloud generation. Analyses that previously took half a day to one day can often be completed in the cloud in tens of minutes to a few hours. Also, because processing is automatically executed with the push of a single button, users do not need to constantly monitor it and can proceed with other tasks in parallel. As a result, the entire project is accelerated and becomes more efficient.

• High-performance PCs not required: Users only need a PC with typical specifications, as heavy computations are handled by the cloud. This reduces the initial investment required to purchase expensive new workstations or to procure software licenses. Because anyone in the company can access it from their own PC, it can be used flexibly regardless of location or device.

• Always have access to the latest technology: Because software updates and feature additions are carried out on the cloud service side as needed, users can always benefit from the latest algorithms and features. There is no need to maintain software or perform version upgrades in-house, and accuracy can improve before you even notice. The ability to automatically receive the benefits of technological innovation is also unique to the cloud.

• Data management and sharing made easy: Output data such as generated point clouds and orthophotos are stored in the cloud. Because large volumes of data can be centrally managed per project, confusion like "Where's that hard drive?" or "Which file is the latest?" is reduced. By granting viewing permissions to stakeholders on the cloud, the same 3D data can be accessed via a browser even from remote locations. There is no need to distribute copies via USB, and all stakeholders can always access the latest data.

• Secure backups: Cloud services typically perform automatic server-side backups and redundancy. Even if a user's PC fails or data is accidentally deleted, you can rest assured if the original remains in the cloud. In addition, security measures such as encrypted communications and strict access controls are implemented, so even highly confidential surveying data can be handled securely.

• A wide range of output options: Depending on the service, you can obtain not only point cloud data but also accompanying deliverables in a one-stop process. For example, some services automatically create an orthomosaic image (an overhead, map-like image created by compositing aerial photographs) at the same time as generating a point cloud from uploaded photos. In addition, features have emerged that let you simply draw an arbitrary line on a cloud-based 3D viewer and instantly generate and download cross-sectional drawing data (DXF, etc.) along that section. Automating tasks that previously required separate software or manual effort, thereby streamlining the entire post-processing workflow after data acquisition, is a major attraction.

As described above, leveraging the cloud enables time and cost savings, improved operational efficiency, and an expanded scope of data utilization. An environment in which advanced 3D modeling can be performed instantly with only photographic data has now become a powerful weapon for operational improvement.

How to Choose a Cloud Service for Point Cloud Generation

Currently, several cloud services support point cloud generation, and each has different characteristics and areas of expertise. When selecting a service that matches your company's needs, it's a good idea to check the following points.

• Ease of use: Verify whether the user interface is intuitive and easy to understand, and whether comprehensive manuals and support are available. If the service offers Japanese-language support, staff who are not confident in English can use it with confidence. It is important that the system allows tasks to be completed simply by uploading photos, even without specialized knowledge.

• Processing speed and supported scale: Even in the cloud, processing times vary by service. It’s a good idea to get a rough sense of processing speed from past performance and user reviews. Check how long it takes to produce results when you feed in on the order of several hundred photos, and whether it can withstand large-scale projects. Also check the maximum data volume you can upload at once and the maximum point-cloud size to confirm the scales it can handle.

• Cost structure：The pricing model is also an important consideration when selecting a service. Many cloud services use monthly subscription or pay-as-you-go billing (pay only for what you use). Consider which pricing plan is economical in light of your organization's usage frequency and project scale. Even if there is no initial fee, we recommend estimating whether the running costs will become too burdensome over the long term.

• Types of output data: Be sure to check the types of deliverables you can obtain from the service. In practice, it's important not only whether point cloud data (PLY or LAS formats, etc.) can be provided, but also whether required formats such as orthophotos, 3D mesh models, and DXF drawings usable in CAD software can be exported. Compatibility with the software your company uses and whether you can obtain the analysis results you need (for example, reports for earthwork volume calculations or equipment inspections) are also points to compare.

• Data sharing features: Some cloud services offer robust platform features that allow generated point clouds to be shared and viewed in a browser. The ability to show point clouds through a 3D viewer simply by giving stakeholders a URL is extremely convenient for meetings with remote participants and for reporting to clients. It's also reassuring to check whether data-sharing security settings (such as password protection or links with expiration) are provided.

• Support and Track Record: If the service is provided by a domestic company, you can expect support in Japanese and implementation assistance. For beginners who are unsure about operation, whether there is a support system that allows easy inquiries by phone or email is an important factor. Also, if past implementation records or case studies are published, they can serve as a reference to see whether the service is being used for purposes similar to your company. Services with a proven track record can provide reassurance in terms of reliability.

Based on the above, compare each service and choose the one that matches your company's use case (for example, whether it's civil surveying or equipment inspection). If you're unsure, trying a trial version or a demonstration first is a good option. By actually using it, you should be able to get a concrete sense of the user experience, the quality of the results, and an estimate of processing time.

Key points for implementation (Tips for success and precautions)

Finally, let's cover the key points to keep in mind when actually deploying and operating a cloud-based point cloud service. Here are some tips and considerations to help you get started smoothly and avoid mistakes.

• Prepare high-quality photographic data: The quality of the source photos determines the accuracy of the point cloud generation results. When shooting, capture sharp, in-focus images and shoot in as bright an environment as possible. For drone aerial photography, establish a flight plan in advance that ensures sufficient overlap (redundancy rate) (a guideline is at least 70% front, back, left, and right). When photographing from the ground, also take photos from various angles around the subject so as to surround it, being careful to avoid blind spots and gaps between images.

• Prepare your network environment: To upload a large number of photo files to the cloud, it is best to have a fast, stable internet connection. When uploading from off-site locations using a mobile router, allow extra time or consider returning to the office and sending the files in bulk. Because data usage will be high, consider using a Wi-Fi environment or subscribing to a large-capacity data plan.

• Test on a small scale: When introducing it for the first time, rather than using it immediately on a large-scale project, we recommend trying it in a small area. Test the entire workflow on familiar structures or within a limited area and check the results; this will reveal the challenges you will face when operating at production scale. By gradually increasing the scale, you can effectively accumulate know-how while keeping the risks from failures down.

• Measures to improve accuracy: If positional accuracy is important for surveying applications, consider not only service-side features but also using RTK-capable devices and ground control points (GCPs) during the capture stage. For example, capturing images with an RTK-equipped drone will tag each photo with highly accurate position data, and the point cloud produced after cloud processing will also have improved coordinate accuracy. Alternatively, placing known points on site and providing them as references during analysis can enable the resulting point cloud to achieve accuracy equivalent to survey maps. Depending on the required level of accuracy, it is advisable to make use of such additional measures.

• Check legal regulations (when using drones): If you will be taking photographs with a drone, compliance with relevant aviation regulations is essential. Beforehand, confirm whether the area falls under a no-fly zone, whether any required permit applications have been completed, and whether the flight requires an operator's license. Including safety management, enforce on-site flight rules thoroughly and take care to avoid accidents and violations.

• Internal communication and training: When introducing new technology, it helps to inform on-site staff and provide simple hands-on training to ensure a smooth rollout. Even if the service is "just take photos and upload them," everyone will be confused at first. Make use of vendor briefings and training videos so that the person in charge becomes familiar with the basic operations. Sharing the support desk contact information is also effective.

By preparing with the above points in mind, you should be able to maximize the effectiveness of introducing a cloud-based point cloud service. With appropriate advance preparations and operations that gradually acclimate personnel, you should be able to steadily advance on-site digital transformation (DX).

Recommendation for Simple Surveying with LRTK

So far, we have explained photo-upload-based cloud point services, and finally I would like to introduce the latest simple surveying tool that leverages them: LRTK (El-Arr-Tee-Kay). LRTK is an innovative solution developed with the concept of "point-cloud measurement completed entirely on a smartphone." By attaching a compact, high-precision GNSS receiver called "LRTK Phone" to a handheld iPhone and simply launching the dedicated app, it is designed so that anyone can easily perform high-precision 3D surveying.

Traditionally, achieving centimeter-level positioning accuracy required specialized GNSS equipment (RTK positioning systems) and skilled surveying techniques. LRTK, by combining a smartphone and a compact RTK receiver, realizes the convenience of allowing a single person to perform field surveying with just one pocket-sized piece of equipment. Actual use is also very simple: you only need to walk around the area you want to survey following the app’s instructions, and the surrounding point cloud data is acquired in real time. By combining the smartphone’s built-in LiDAR scanner with RTK’s high-precision positional information, it becomes possible to quickly obtain detailed point clouds capturing everything from terrain to structures.

Furthermore, data acquired with LRTK can be put to immediate use by integrating with cloud services. If you upload the point-cloud data recorded on-site or the photos taken by drone directly to the LRTK Cloud, a point-cloud model with positional coordinates is generated automatically, and it can be edited and shared in the browser. Because you don’t need a high-spec PC of your own, it offers the flexibility to handle data consistently from small surveys to large-scale projects. LRTK, which lets you try cutting-edge 3D technology while keeping initial costs low, is a tool well suited to the on-site DX approach of starting small and gradually expanding your scope of use. Especially for beginners who are interested in leveraging point clouds but have not been able to take the first step, why not use LRTK to easily experience 3D point-cloud surveying?

FAQ

Q: Can point cloud data be created without special equipment? A: Yes. By using photos taken with a smartphone camera or a commercially available drone, you can generate point cloud data without a special laser scanner. With cloud-based photo analysis services, you can create 3D models simply by uploading photos, even if you don't have a high-performance PC on hand. However, photo quality and the number of images affect the quality, so it's important to capture as many clear images as possible.

Q: Can beginners master it? A: Many cloud-based point cloud services are designed with beginners in mind, and you can get results simply by following basic procedures. Many services offer Japanese-language interfaces and manuals, so advanced specialized knowledge is not required. If you are uneasy about the operation, choosing domestic services with comprehensive support desks will allow you to learn while asking questions when you run into problems.

Q: Are there any disadvantages to using the cloud? A: A point to note is the dependence on your internet connection. When uploading a large number of photos, a slow connection can mean processing takes longer and communication costs may increase. Also, fees for cloud services are ongoing. However, as mentioned above, considering the advantage of not needing to purchase expensive PCs or software, in many cases the overall cost benefits are substantial. For highly confidential data, choose a proven, reliable service and check the terms of service, etc., to confirm how your data will be handled.

Q: What level of accuracy can be achieved for point clouds? A: The accuracy of point clouds from photogrammetry varies depending on shooting methods and equipment used, but generally planar (horizontal) errors are on the order of a few centimeters (a few in), and vertical errors are around 10 cm (3.9 in). Using high-precision RTK-enabled drones and a sufficient number of ground control points (GCPs) can achieve accuracy within a few centimeters (within a few in). However, accuracy tends to decrease in areas with dense vegetation or on surfaces with few distinctive features, such as water. Depending on the required accuracy, consider adjustments to photography, combining RTK/GCP, or, where appropriate, using LiDAR surveying.

Q: I'm unsure which cloud service to choose. What should I do first? A: I recommend first trying the trial or demo of the services you're interested in. By processing your own captured data, you can concretely assess the service's usability, the quality of its results, and its processing speed. You can also get a sense of the provider's support responsiveness. On top of that, check the points discussed in this article (usability, speed, cost, output formats, etc.) and evaluate whether they match your company's use case. By trying and comparing multiple services, you'll naturally see which one is easiest for your team to use.