How to fix point clouds that are too heavy to open! 5 tips to speed up visualization | PC requirements summarized

Table of Contents

• Causes of point cloud data being too heavy to open

• Tip 1: Save point cloud files in a compressed format

• Tip 2: Split the data by range and type

• Tip 3: Reduce size by thinning out unnecessary points and excessive density

• Tip 4: Optimize display settings and use a dedicated viewer

• Tip 5: Leverage cloud services to distribute the load

• PC requirements for handling heavy point clouds

• Summary

3D point cloud data acquired on construction, civil engineering, and surveying sites has been increasingly utilized in recent years. However, many field personnel may be troubled by issues such as "the point cloud file is too large to open" or "it takes too long to load." High-density point clouds can cause file sizes to balloon, and in some cases a computer may freeze or the software may crash.

In this article, we unravel the causes of the problem of "heavy point clouds not opening" and explain five tips for speeding up point cloud visualization. We also outline the PC specs (PC requirements) needed to handle large datasets. By applying these measures, you will be able to use 3D point clouds smoothly without being troubled by large point cloud data.

Causes of point cloud data being too large to open

There are several reasons why point cloud data can become large. First is the sheer number of points themselves. Because a point cloud consists of countless points that make up a space, when the acquisition area is wide or the measurement accuracy is high, the number of points can increase explosively and the file size can balloon. For example, if you laser-scan at high density an area several hundred meters (several hundred ft) on a side, the total number of points can reach tens of millions, and file sizes commonly range from several hundred MB to, in some cases, several GB. This is an amount of information orders of magnitude greater than traditional 2D drawings or photographic data.

Second, the amount of information each point carries is also a cause of increased storage. Typically, each point in a point cloud contains not only X, Y, and Z coordinates but also attribute information such as RGB color, return intensity, and classification codes. For example, in airborne LiDAR surveys, the acquired point cloud may be assigned classification labels such as ground, water surface, vegetation, and buildings, as well as color values based on aerial photographs. These attribute data are useful for analysis and visualization, but they increase the data volume per point and thus contribute to larger overall file sizes.

Furthermore, the fact that the range of targets (ground objects) contained in point cloud data is extensive should not be overlooked. If, instead of just the terrain, everything on site—buildings, trees, structures, and so on—is converted into a point cloud together, the dataset will include a large number of points beyond the parts actually needed, unnecessarily making the data heavy. Even in cases where, depending on the intended use, "knowing only the terrain of the ground surface is sufficient," if the source data contains a mixture of buildings and trees, you will end up handling files that are heavier than necessary.

Finally, the data storage format is also important. Saving point clouds in text formats (such as CSV or PTS) is very inefficient because numeric values are recorded as strings, and for the same content the file size becomes many times larger than with binary formats (such as LAS). Furthermore, if you compress LAS files using a dedicated algorithm into LAZ format, the size can be reduced to about 10–20% of LAS (that is, roughly 1/10). Text formats have the advantage of being human-readable, but they can lead to the counterproductive situation where files become so huge that they cannot be opened in a text editor. Overall, the sheer number of points, the large amount of information per point, and the inefficiency of the data format combine to make point cloud files easily become "too heavy to open".

Tip 1: Save point cloud files in a compressed format

To reduce the size of the point cloud data itself, reviewing the file format is essential. Especially effective is storing point cloud files in compressed form. The LAS format, the standard for point cloud data, is binary and efficient, but converting it to the specialized compressed LAZ format can reduce data size to about one-tenth of the original (around 10–20%). Moreover, compression to LAZ is lossless (reversible), so when decompressed the original data can be fully restored. Because files can be made smaller without losing any information, LAZ has become so widespread that it is now almost universally used for distributing and sharing LAS files.

If your point cloud data is stored in a text format (such as CSV or PTS), simply converting it to a binary format like LAS can significantly reduce file size. Text formats contain a lot of redundant character information, so the same point cloud can balloon to several times or more than ten times the file size. Converting to LAS/LAZ is not a difficult operation. Using free conversion tools or features in point cloud software, you can batch-convert with a single command or a few clicks. With compressed LAZ files, data loading times are shortened and network transmission becomes smoother, making the data much easier to handle.

Tip 2: Divide data by range and type

Another way to reduce the size of point cloud files and make them easier to handle is to split and manage the data appropriately. If a vast area is consolidated into a single file, it is advisable to divide the area into tiled files by area. For example, dividing the area into grids of 1 km square (3280.8 ft) or 500 m square (1640.4 ft) will keep each file's size to a few hundred MB. In fact, the Geospatial Information Authority of Japan provides nationwide lidar point cloud data split into LAS/LAZ files by 1:25,000 map tiles. By separating files by area, you only need to load the files covering the range required for a task, which greatly reduces the load on your PC. Also, including the bounding coordinates or mesh numbers in the filenames makes it immediately clear which data corresponds to which region, simplifying team management.

It can also be effective to split point clouds by the "type" of points they contain. Point cloud data contains a mix of features such as ground surface points, building points, and tree points. By separating the data by point type in advance, you can extract and work with only the information needed for each purpose. For example, for terrain analysis you would manage a dataset consisting only of ground (surface) points, while if the goal is structural modeling you would manage a dataset composed only of building and structure points. Recent point cloud processing software and cloud services have robust AI-based automatic classification features, making it easy to automatically label and separate layers such as ground, low vegetation, high vegetation, and buildings. Even with existing LAS data, if points have classification codes (Class) assigned, you can use filter functions to extract specific classes and write them out to separate files.

In this way, by appropriately segmenting and managing point clouds by area or by content (types of features), you can avoid situations such as "too heavy to open" or "wasting time by loading unnecessary parts." After splitting, it is also important to organize the resulting files with clear naming conventions and folder structures so that stakeholders are not confused about which data to use. In addition, advanced approaches such as Entwine Point Tiles (EPT), which hierarchically tile point clouds based on spatial indexes so the viewer loads only the tiles for the required extent and resolution, have been put into practical use. When distributing or sharing large-scale data, it is advisable to consider utilizing such approaches.

Tip 3: Thin out unnecessary points and excessive density to reduce weight

Reducing the number of points in the data itself is also a basic method for lightening display. Point clouds acquired on site may contain points or noise that are not very meaningful for analysis. For example, isolated points caused by airborne noise or measurement errors, or points that were captured redundantly. By removing such unnecessary points in advance, you can reduce data volume while improving quality.

Furthermore, if the point-cloud density is excessively high, you can lighten the dataset by appropriately thinning it to reduce the number of points. Especially for uses that do not require a high level of detail, you can often achieve satisfactory results without using the full number of points from the original data. Many point-cloud processing software packages include thinning (sampling) functions that can reduce point-cloud density by, for example, randomly dropping a fixed percentage of points or deciding intervals on a grid and leaving one representative point per cell. For example, by setting a grid interval and keeping only the single most representative point in each cell, you can greatly compress the data volume while preserving the apparent shape.

However, indiscriminately reducing the number of points can risk losing important details. When performing thinning (downsampling), it is important to back up the original data and determine the number of points that is necessary and sufficient for your analysis objectives. Point cloud data from which excessive duplicate data and obvious noise have been removed and that has been adjusted to an appropriate density will have a smaller file size, faster loading, and smoother visualization and analysis.

Tip 4: Adjust display settings and use a dedicated viewer

To make point cloud data display more responsive, software-side settings and the choice of tools are also important. If the point cloud viewer or CAD software you use has display options, adjusting them can improve performance. For example, setting an upper limit on the number of points to display (point budget) and limiting the number of points drawn at once can avoid excessive load. Some software includes an LOD (Level of Detail) function that automatically thins points and lowers display resolution when zoomed out, so it’s useful to take advantage of it. Also, if there are "region filter" or "crop" functions that display only the necessary area and hide unnecessary parts, you can switch between a full view for an overview and a local view for detailed inspection, avoiding drawing unnecessary points and keeping operation smooth.

If the software you’re using renders large point clouds too slowly, consider using a faster dedicated viewer or a cloud service. There are lightweight viewers that can run smoothly even with point clouds on the order of hundreds of millions of points, as well as online tools that let you work with point clouds in a web browser. Those tools are designed with techniques such as spatial indexing and visibility-range tile loading, so they can render only the needed parts on the fly without loading the entire dataset into your PC’s memory. As a result, you can comfortably view and share very large datasets even with limited PC specs. However, importing data into another tool does require some effort, so it’s wise to choose based on the scale and purpose of your project.

Furthermore, a PC's graphics performance also affects display speed. Many programs that perform 3D rendering make use of a GPU (graphics card), so rendering may be smoother on a machine with a dedicated GPU. On PCs with only integrated graphics, rendering can take longer, so if possible it is preferable to work with point clouds in an environment that has a graphics card. Some software allows you to enable GPU support in the rendering engine settings, so check the configuration options.

Tip 5: Leverage cloud services to distribute load

In recent years, methods that leverage cloud technology to reduce the burden of point cloud processing have also attracted attention. The idea is that you do not necessarily have to process and display large point cloud datasets solely on your local PC; you can entrust them to powerful servers on the Internet. If you upload data to a dedicated point cloud cloud service, the server can convert it into a format optimized for viewing and run heavy analytical processes in the background. Users can check results via a web browser or a dedicated app, or stream only the required areas, making it possible to handle large-volume point clouds smoothly regardless of their local PC specs.

The benefits of using the cloud are not limited to simply solving PC load issues. By centrally managing point cloud data online, multiple stakeholders can view and edit it simultaneously and more easily share always up-to-date data. There is no need to transfer huge files via USB flash drives or external HDDs, and it prevents confusion caused by version mismatches. In addition, point clouds captured on-site can be uploaded to the cloud on the spot so office technicians can check them immediately, enabling real-time collaboration. Previously there was a time lag of "measure on-site → take it back for processing → share results," but by using the cloud the field and the office connect seamlessly, allowing immediate use of 3D data.

By using cloud services effectively, you can avoid being overwhelmed by heavy point cloud data and build a flexible workflow that isn’t constrained by your PC’s performance. Depending on your company’s security requirements and network environment, if the point clouds you handle are extremely large—on the order of several billion points—it may be worth considering cloud usage as an option.

PC requirements for handling large point clouds

Even with the data-reduction and cloud-utilization measures described above, there are situations where you realistically have to handle large point clouds on a local PC. In those cases, the PC’s performance (specs) is indeed a crucial key to displaying and processing point clouds smoothly. So, what level of PC specs is required to handle heavy point clouds?

The key components are the CPU, memory, and GPU (graphics). First, for the CPU, a high-end model with a high clock speed and multiple cores is desirable. CPU power makes a difference when loading, transforming, and analyzing point clouds, so if possible you should consider a creator-oriented or workstation-class processor rather than a typical office PC CPU.

Next, memory (RAM) is the most important resource for a point-cloud processing PC. Point cloud data are loaded into memory when opened, so if the installed memory is insufficient, applications may crash or the entire OS may freeze. Small point clouds can run on around 16 GB, but for comfortable performance you should aim for 32 GB or more. In large-scale projects, it is not uncommon to equip large-capacity memory such as 64 GB or 128 GB. Especially for photogrammetry (photo surveying) software, some recommend 128 GB-class memory for generating point clouds that exceed tens of millions of points. If you regularly work with point clouds, it is best to install as much extra memory as possible.

And the GPU (graphics board) is also important. In 3D point cloud visualization, because millions to hundreds of millions of points are rendered simultaneously, a powerful GPU that accelerates rendering will make the display noticeably smoother. High-performance discrete GPUs for gaming and CAD offer overwhelmingly better rendering performance than integrated graphics, and if paired with large-capacity memory they will allow you to rotate and zoom heavy point cloud data without stress. The larger the GPU’s onboard memory (VRAM), the more points can be handled at once, so if possible you should equip a mid- to high-end graphics card.

Also, choosing a fast storage device (storage) is effective. Compared to HDDs, SSDs — and preferably NVMe-connected SSDs — can greatly shorten the load times of point cloud files that can reach tens of GB. Because the data itself tends to be very large, always keep ample free disk space and consider using external drives or a NAS as needed. Also, using 64bit versions of the OS and software is essential (a 32bit environment limits the amount of memory that can be handled). Most modern point cloud processing software is 64bit-compatible, so if you are still using an old PC, updating your environment can be expected to improve performance.

In summary, to handle heavy point cloud data locally, the ideal combination is a "high-performance CPU + large-capacity RAM + high-speed GPU + high-speed storage (SSD)". With a machine equipped with these components, combined with the aforementioned data-reduction measures, you should be able to work smoothly with fairly large datasets. Conversely, forcing work on an underpowered PC can lead to long waits just to load data, or choppy responsiveness that prevents productive work. If you plan to use point clouds seriously in a business setting, consider investing in your hardware and working environment.

Summary

In this article, we introduced five tips for speeding up visualization as measures to address point cloud data that are too large to open. By optimizing file formats, splitting and decimating data, using appropriate tools, and leveraging hardware and the cloud, large point clouds that were previously difficult to handle should become far easier to work with. We also explained PC requirements, and in short, a system equipped with a high-performance CPU, large-capacity memory, a fast GPU, and an SSD will make handling them much simpler.

More recently, new solutions have emerged that connect the field directly to the cloud to enable real-time utilization of point cloud data. A representative example is the high-precision GNSS positioning device "LRTK" that can be attached to an iPhone. LRTK is a compact device that, used in combination with an iPhone, achieves centimeter-level high-precision positioning (half-inch accuracy), and can provide accurate position information to point clouds and survey data collected on site in real time. For example, at a civil engineering site, simply walking around the construction area with a smartphone fitted with LRTK allows you to measure the surrounding terrain and structures as point clouds on the spot, and all acquired point clouds are immediately assigned world coordinates. After measurement, the data can be uploaded to the cloud from the smartphone with a single tap, making it possible to share results immediately from the field without returning to the office. Uploaded point cloud data is automatically organized and visualized in the cloud, and all stakeholders can view and verify it via a browser. There is no need to exchange heavy data individually, and the time lag between the field and the office is greatly reduced.

By adopting a cloud-integrated workflow that leverages LRTK in this way, you'll encounter fewer situations troubled by "point clouds that are too heavy to open." An environment is taking shape in which anyone can easily acquire and share high-precision 3D data, and the future of "heavy PCs are no longer necessary" is becoming increasingly realistic. If you've been hesitating to use point cloud data because of its large size, consider embracing the power of the latest technology and taking the next step in point cloud utilization.