A groundbreaking method to automatically extract a list of survey points from point cloud data and output CSV

Table of contents

• Introduction

• What point cloud data and a list of survey points are

• Conventional methods of obtaining survey points and their challenges

• Approaches to automatically extracting survey points from point clouds

• Benefits of automatic extraction technology

• Steps and advantages of CSV output

• Features and convenience of simple surveying with LRTK

• FAQ

Introduction

In recent years, methods for acquiring point cloud data—such as 3D laser scanners and drone photogrammetry—have rapidly spread across the construction, civil engineering, and surveying fields. Point cloud data consist of numerous measured points (coordinates) that make up an object or terrain, and they are characterized by the ability to digitally record site conditions with high precision. However, extracting the information needed for practical work from 3D point cloud data containing millions to hundreds of millions of points is labor-intensive. In particular, obtaining the traditionally important “list of survey points” (a coordinate list of observation points) from a point cloud requires manually identifying each required point and recording its coordinates one by one, which is highly inefficient.

Now, leveraging automation technologies for point cloud data analysis, a revolutionary method that extracts only the necessary survey points from point clouds and outputs them as a CSV-format coordinate list is attracting attention. This article explains the background, benefits, and concrete approaches to automatically extracting a list of survey points from point cloud data and outputting it as CSV. Doing so makes it possible to handle large amounts of point cloud data more efficiently and greatly reduce the time required to produce surveying deliverables. Such automation technologies are expected to contribute significantly to productivity improvements in surveying operations and to DX (digital transformation).

What point cloud data and a list of survey points are

Point cloud data are collections of three-dimensional coordinate data for many points. They are acquired by laser scanners or photogrammetry (such as SfM) and represent surfaces of terrain and structures as innumerable points. Each point carries (X, Y, Z) position information, and by increasing point density, the site can be digitally reproduced almost exactly as it is. While point clouds can record terrain relief and the shapes of structures in detail, they also tend to be very large in size as-is, making handling and analysis cumbersome.

On the other hand, a list of survey points is a list of coordinate points obtained by surveying. Typically, for each observed point the ID, point name, and coordinate values (X, Y, Z) are compiled in a tabular format and submitted or stored as a deliverable in CSV or spreadsheet file formats. Especially in public surveying, a coordinate table of observation points is often required as a deliverable, and if output automatically in CSV format, preparing such reports becomes easy. For example, in civil engineering projects, points indicating ground elevation and key coordinates of structures are organized as a survey point list and imported into design drawings or CAD software for use. A survey point list is a compact dataset containing only the necessary points, suitable for stakeholders to check coordinates or use them in subsequent calculations. However, in conventional surveying, creating this list required measuring and recording each point with equipment one by one, which demanded a great deal of time and effort to cover wide areas.

Conventional methods of obtaining survey points and their challenges

Traditionally, when conducting topographic surveys or as-built measurements, surveyors observed survey points on-site one point at a time and recorded their coordinates. Using total stations or GNSS survey instruments, important points were selected and measured, and when there were many points, it could take several days to cover the whole area. With this method, only points judged necessary in advance can be obtained, so if you later realize “we should have measured another location,” you must return to the field. Also, in locations with poor footing or hazardous areas, it may be impossible to take survey points, making complete coverage of site conditions difficult.

By contrast, acquiring point clouds with 3D scanners or UAV photogrammetry yields high-density point data of the entire site in a short time. This allows recording details that were previously missed, but it also creates a new challenge: “extracting the necessary information from the point cloud.” Point cloud data are not automatically organized by feature or object type, so to obtain information that used to be acquired by conventional methods (for example, ground elevation or the corner positions of structures), humans must visually inspect the point cloud, select the corresponding points one by one, and read off their coordinates. Searching through vast numbers of points to find target points is extremely tedious and requires specialized software operation skills and a significant amount of time. In other words, while data acquisition efficiency has drastically improved, new burdens have arisen in the subsequent data processing.

Approaches to automatically extracting survey points from point clouds

Automatically extracting survey points from point clouds leverages advanced algorithms for point cloud processing. First, preprocessing such as noise removal and coordinate transformation is performed to prepare the data for analysis. Next, features such as ground surfaces, planes, and edges of structures within the point cloud are recognized to narrow down candidate survey points.

For example, if the purpose is topographic surveying, ground points can be classified from the point cloud (ground surface extraction), and representative points can be selected at regular intervals to automatically generate a set of survey points representing surface relief. If you want to extract corners or edges of structures, plane detection algorithms (such as RANSAC) can identify wall or floor planes and compute corner points at their intersections. In recent years, advances in AI have enabled solutions that automatically classify objects in point clouds (ground, buildings, trees, etc.) or detect feature points (such as pole tops or edge lines). By combining these techniques, it is possible to mechanically extract and list “the required survey points” according to conditions specified in advance by a human.

In actual automatic extraction workflows, the user only needs to set the types of points and criteria to extract (for example, “extract contour points of the ground at ○ m (○ ft) intervals” or “extract the tops of artificial structures”), and the software performs the remaining processing at high speed. This process of picking only the points of interest from vast point clouds is executed at speeds and with consistency unattainable by manual work, eliminating subjective variation.

Benefits of automatic extraction technology

Automatic extraction technology offers many advantages that manual methods do not. The main points are as follows:

• Significant improvement in work efficiency: Because computers perform rapid point cloud analysis, the time required to create a list of survey points can be greatly reduced. In one case at a major construction company, measurements that had taken 45 minutes per cross section for checking bridge components were reduced to 20 minutes after introducing automation technology, more than doubling efficiency. In another project, analyzing terrain point clouds obtained by drone to calculate earthwork quantities reduced what had previously taken nearly a week of cumulative work to about one day, achieving an efficiency improvement of more than tenfold in terms of person-hours. By utilizing automatic extraction, personnel and time required for surveying and analysis can be dramatically reduced, shortening construction schedules and cutting costs.

• Improved data comprehensiveness: Since point cloud data contain information about every corner of the site, locations that were previously overlooked or later needed can be extracted as long as the data exist. Recording the entire site in a single measurement prevents rework due to “having to measure again,” ensuring all necessary information is captured. Furthermore, point clouds can non-contact measure hazardous or inaccessible areas, enabling both improved safety and richer information.

• Ensured quality and consistency: Automatic processing selects all survey points according to consistent criteria, eliminating variation due to the experience or subjective judgment of personnel. Reproducibility is assured—if processed under the same conditions, anyone will obtain the same results—which also reduces human error. In terms of accuracy, if the point cloud itself was acquired with high precision, automatically extracted coordinates can have accuracy comparable to manual surveying (in fact, there are reports where dimensions obtained from point clouds matched those from traditional direct measurements). Thus, automatic extraction technology not only improves efficiency but also has great potential to surpass conventional methods in data quality and site safety.

• Labor savings and effective use of human resources: Automation enables a small number of people to handle wide-area surveying work, significantly addressing labor shortages that are common at construction sites today. Experienced technicians are freed from tedious repetitive tasks and can focus on higher-value analysis and decision-making. Reducing the time people spend on-site also helps ease safety management burdens and contributes to work-style reforms.

Steps and advantages of CSV output

Automatically extracted survey points can be output in CSV format. CSV (Comma-Separated Values) is a simple text-based data format ideal for handling a list of survey points. Many point cloud processing and analysis tools include functions to export selected point cloud data as CSV files, allowing you to save a coordinate list with a single click after extraction completes.

A CSV output file typically records each survey point’s ID or name, X coordinate, Y coordinate, Z coordinate, and attribute information as columns as needed. Anyone can open this file in a spreadsheet application to immediately check coordinate values, and it is also easy to import it into CAD or GIS software to plot points on a map. Because it is text-based, the file size is small, making it easy to share by email or via the cloud. Even if only a laptop is available on-site, CSV allows checking content without a special viewer—this universality and immediacy are major advantages of CSV output.

CSV also serves well as an intermediate format. If you export coordinate data extracted from point clouds to CSV, you can later convert it to other formats as needed. For example, creating DXF drawings for design or converting to LandXML for consistency with a 3D model can proceed smoothly if you have a base coordinate CSV. Conversely, you can first output CSV from the point cloud for quick on-site checks and later re-export to more precise formats when detailed drafting is required. Making CSV the initial output expands the possibilities for subsequent processing and utilization.

Features and convenience of simple surveying with LRTK

Recently, RTK (real-time kinematic) positioning technology has become more accessible and easier to use. A representative approach is known as LRTK. LRTK (※) combines a high-precision GNSS antenna and a dedicated app with easy devices such as smartphones, enabling RTK surveying—which previously required specialized equipment—to be used as simple surveying by anyone. By using LRTK, you can achieve centimeter-level high-precision positioning (half-inch accuracy) with just a smartphone, without owning expensive dedicated surveying instruments, allowing you to obtain position coordinates on-site whenever needed.

Real-time correction services and cloud integration have automated the formerly complex tasks of communicating with base stations and configuring data, so even technicians without surveying expertise can perform positioning intuitively. Moreover, acquired positioning data are immediately saved and managed in the cloud, allowing records to be kept without worrying about file formats on the spot. You can later export the data as CSV coordinate lists or map data as needed, making data utilization flexible.

Thus, simple surveying using LRTK is attracting attention as a new method that delivers affordable, fast, and easy-to-handle high-precision positioning. It can be powerful in a wide range of scenarios, from everyday minor surveying needs to rapid on-site situation assessment during disasters. If conventional RTK equipment feels costly or difficult to operate, LRTK lowers the barrier to entry and is worth trying. By utilizing the latest technologies, the workflow from acquiring to using surveying data is becoming dramatically more efficient, and an era is approaching in which necessary information can be put to immediate use regardless of format differences. This aligns with the Ministry of Land, Infrastructure, Transport and Tourism’s i-Construction initiative and is a technology likely to further promote the digitalization of surveying.

(※LRTK: a generic term for a form of real-time kinematic positioning usable with simple devices, not tied to any specific brand)

FAQ

Q1. Do I need expensive specialized software to output a list of survey points from point cloud data as CSV? A1. No, you do not necessarily need expensive software. Basic point cloud-to-CSV conversion can be handled by free point cloud viewers or open-source tools. For example, many programs can output point clouds as text files of XYZ coordinates. Recently, cloud services that process point clouds and allow CSV downloads of results have also appeared. More advanced features that include automatic extraction may require specialized tools, but it’s good to start with easily handled tools.

Q2. Can the accuracy of automatically extracted coordinate values be trusted? A2. Yes, high accuracy can be obtained if proper procedures are followed. The accuracy of automatic extraction depends on the accuracy of the source point cloud. If the point cloud was acquired with high precision and the coordinate system is accurately aligned, the extracted survey point coordinates will generally be reliable. There are cases where dimensions and coordinates computed from point clouds were comparable to traditional manual measurements. However, for important survey points used as final deliverables, it is preferable to perform additional on-site verification when possible. For example, measuring some extracted results with a total station and comparing them can confirm there are no errors. It is advisable to perform spot double-checks rather than relying solely on automatic processing.

Q3. Do I need a high-performance PC to process point cloud data? Can large point clouds be handled? A3. Since point cloud data tend to be large, it is true that a reasonably high-performance PC and sufficient memory are desirable. However, recent software has features for handling large data, and it is increasingly possible to process point clouds with tens of millions of points on typical business PCs. For example, dividing point clouds into tiles for loading or processing only the required area sequentially can reduce PC load. Also, using cloud-based point cloud processing services allows large data to be analyzed regardless of local PC performance. By choosing the appropriate method for the requirements, processing performance is unlikely to be a bottleneck except for extremely large point clouds. With a proper environment, large-scale point cloud processing is practically feasible.

Q4. If automatic extraction exists, is conventional on-site surveying no longer necessary? A4. Automatic extraction and point cloud surveying are powerful tools, but conventional methods remain important in some situations. Point clouds cannot capture areas that are not visible (such as behind occlusions or underground), so additional on-site measurements may be required. Also, to place extracted results into an accurate coordinate system, traditional tasks like establishing control points and georeferencing are still indispensable. Furthermore, interpreting automatic extraction results and making final judgments are human responsibilities. In short, automation does not solve everything on its own—combining it with human expertise is essential. Using the latest technologies while performing checks by human eyes and experience at key points will yield more reliable deliverables.

Q5. How should point cloud data be acquired? Do I need expensive specialized equipment? A5. No, you do not necessarily have to purchase expensive equipment. For example, aerial photography with a camera mounted on a drone and processing the photos into 3D point clouds using photogrammetry software is a relatively low-cost method that provides wide-area data and has become widespread. Also, recent smartphones and tablets include simple LiDAR sensors, which can scan point clouds easily for small areas. For large projects, you can rent high-precision laser scanners or outsource point cloud acquisition to specialized survey companies. As technology advances, the barriers to acquiring point cloud data continue to fall, and it is becoming possible to use 3D surveying methods appropriate to your situation even without owning advanced equipment. Going forward, 3D point clouds will increasingly become a routine surveying method rather than something exceptional.