6 Steps to Create Point Cloud Data for a Solar Power Plant Using Drone Surveying

At solar power plants, a large site contains numerous panels, mounting racks, access paths, drainage facilities, slopes, fences, and other elements. It is difficult to grasp the overall picture from ground-level inspections alone, and it can take time to understand changes in terrain and the elevation relationships around equipment. One method used to address this is to create point cloud data based on images and positional information obtained from drone surveys.

Point cloud data serves as fundamental material for confirming the on-site shape in three dimensions. At solar power plants, it can be used to check the terrain undulations of development sites, drainage slopes, slope conditions, heights around mounting racks, deformations of maintenance paths, and topographic changes after disasters. However, simply flying a drone and taking images does not automatically produce point clouds that are practical for fieldwork. It is important to sequentially prepare the objective setting, flight plan, control points, shooting conditions, processing settings, and result verification.

Table of Contents

• Organize the objectives for generating point cloud data at solar power plants

• Step 1: Decide the scope and the items to be checked

• Step 2 Conduct flight planning and safety checks

• Step 3 Prepare reference points and shooting conditions

• Step 4: Capture the site with a drone

• Step 5 Generate point cloud data from images

• Step 6: Confirm the results and apply them to on-site operations

• Common pitfalls that lead to failure when creating point cloud data

• Summary

Clarifying the objectives for generating point cloud data at solar power plants

The purpose of producing point cloud data from drone surveys of solar power plants is not simply to create visually appealing three-dimensional data. What matters in practice is organizing the plant’s condition so it can be checked safely, over a wide area, and at a later time. Inspecting the site on foot tends to bias information toward what is visible, and it is not easy to recheck the same location under the same conditions. If point cloud data is retained, it becomes easier to review heights, distances, and terrain trends at the office after the on-site inspection.

In solar power plants, not only the panels themselves but also the surrounding ground, drainage, slopes, maintenance access roads, and conditions along fence lines are related to the plant’s operations and maintenance. When checking after heavy rain whether drainage routes have changed, whether parts of slopes show signs of deformation, whether ruts or subsidence have occurred on access routes, or whether scouring has occurred in the ground around mounting structures, it can be difficult to judge from planar photographs alone. Using point cloud data allows confirmation of the site’s shape including the vertical dimension, which helps in detecting changes.

On the other hand, point cloud data is not infallible. Areas such as beneath solar panels, ground surfaces covered by trees or grass, highly reflective surfaces, and locations with deep shadows can produce noisy or missing data depending on acquisition conditions. Also, a higher point density does not necessarily make the data easier to use in practice. If you generate data without considering the accuracy, scope, and update frequency required for what you need to verify, you will only increase data volume and make it harder to identify the points you should be looking at.

Therefore, the first thing to consider is, "What is the purpose of creating point cloud data?" The flight area and imaging method will vary depending on whether the goal is to understand the current condition of the entire power plant, to check changes after a disaster, to compare conditions before and after construction, or to assist inspections of drainage and slopes. By clarifying the purpose before planning the work, the point cloud data becomes an easy-to-handle resource for the site.

Step 1: Define the scope and items to be checked

The first step is to decide the drone survey’s target area and the items to be checked. Solar power plants have large sites, and trying to capture the entire area at high density at once increases the number of photos, processing time, and verification work. Rather than capturing everything at the same level of detail each time, dividing the site into an area for an overall overview of the plant and areas you want to focus on for dense point cloud generation makes it easier to create a plan suited to practical work.

For example, if you want to check the overall layout and topography of a power plant, set a fairly wide area that includes the site boundary, panel rows, maintenance walkways, drainage channels, balancing ponds, slopes, and access roads. For inspections after a disaster, make low-lying areas where rainwater tends to collect, confluences of drainage channels, the toe of slopes, the boundaries of cut-and-fill from earthworks, and locations where deformation has been previously observed the primary focus. For post-construction verification, concentrate on the earthwork surfaces, racking installation areas, access routes, and interfaces with surrounding structures.

It's important to articulate the items to be checked in advance. Simply saying "I want to see the terrain" does not determine the required data granularity. Whether you want to examine ground settlement trends, check drainage flow, understand slope surface irregularities, or confirm the height relationship between panel rows and the ground will affect the capture angles, image overlap, and placement of reference points. Sharing the items to be checked among site personnel, maintenance personnel, and surveying personnel can reduce rework in later processes.

Also, confirm who will use the point cloud data and how. Whether the surveyor will only check details in a 3D visualization, whether it will be shown to the client or managers as explanatory material, whether it will be compared with past data to measure changes, or whether it will serve as supplementary material for plan views and cross-sections will change the format and organization of the deliverables. Rather than just handing over the point cloud as-is, planning on producing orthophotos, elevation models, cross-sections, and annotated materials as needed will result in deliverables that are easier to use later.

At this stage, site constraints are also identified and organized. Confirm that there are facilities currently generating power, restricted access areas, electrical and communication equipment, movement routes for work vehicles and inspection personnel, and nearby houses and roads. In solar power plants, safety considerations are essential even in wide-open spaces. Defining the target area is not only important for the efficiency of data acquisition but also serves as a prerequisite for safety management.

Step 2: Conduct flight planning and safety checks

Once the target area is determined, the next step is to prepare the flight plan and conduct safety checks. In drone surveying, you need to decide in advance the flight path, flight altitude, imaging interval, direction of travel, spare batteries, takeoff and landing locations, and emergency procedures. At solar power plants, because panels are widely arranged, a simple grid flight can sometimes capture the entire site easily, but when including slopes, surrounding equipment, fence lines, drainage channels, and so on, it is important to plan to prevent any missed coverage.

Flight altitude affects the resolution of the point cloud and the imaging coverage. Flying lower makes it easier to obtain detailed images, but increases the number of shots and requires attention to distances from obstacles. Flying higher allows efficient imaging of a wider area, but may not be suitable for checking fine details. In point cloud generation for solar power plants, it is effective to separate flight conditions into ranges that prioritize overall understanding and ranges intended for inspecting fine anomalies.

When conducting safety checks, confirm in advance flight rules such as the Aviation Law, restrictions on flight locations, required permits or approvals, and the facility manager’s rules. When flying unmanned aircraft of a certain size or above outdoors, confirmation of aircraft registration, flight procedures, and airspace is required. Even within a power plant’s premises, flight permission and operating conditions may change depending on surrounding roads, residences, transmission equipment, emergency-use airspace, or third-party access.

On site, we check overhead lines, poles, communications equipment, surveillance cameras, fences, trees, adjacent facilities, the positions of workers, and vehicle entry and exit. Because reflections from solar panels can change visibility, we ensure measures are in place so the pilot and assistants do not lose sight of the aircraft. In locations prone to strong winds or gusts, such as valley terrain or open developed sites, the wind felt at ground level can differ from the wind aloft.

When working inside a power plant, coordination with the facility manager is also important. Confirm the flight date and time, access areas, placement of personnel, precautions near electrical equipment, and emergency contact information. If maintenance inspections, mowing, or construction work overlap, adjust the timing so drone operations and ground work do not interfere with each other. Ensuring the overall safety of the operation takes precedence over the accuracy of the point cloud data.

Include the time of day for shooting in the plan. Solar panels are highly reflective, and the way they appear in images changes with the time of day. During periods with extreme backlighting or deep shadows, image processing may have difficulty detecting feature points. Overcast skies are not always ideal, but when the overall image brightness is stable, processing can be easier. Consider the weather, the sun’s direction, and how shadows fall; when making periodic comparisons at the same solar power plant, it is desirable to shoot under conditions as similar as possible.

Step 3: Set up reference points and shooting conditions

To use point cloud data in field operations, it is necessary to standardize how position and height are handled. If you are only generating a point cloud from images, you can reproduce three-dimensional shapes by processing the captured images. However, to compare with drawings or past data, or to verify distances and heights, it is important to properly establish reference points and validation points. If this is left vague, a point cloud that looks tidy can become a deliverable that is difficult to use for on-site decision making.

Reference points are markers used to align point cloud data with on-site coordinates. At solar power plants, choose locations that are easy to identify in imagery and can be installed safely, such as between rows of panels, along maintenance walkways, on prepared surfaces, and near fences. On large sites, if reference points are concentrated on only one side of the area, overall alignment may remain uncertain. Place them so they surround the area of interest, and in locations with elevation differences, also consider verification in the vertical direction.

Validation points are also important. Control points are the points used in processing, while validation points are used to verify how well the processing results match the actual site. If you produce deliverables using only control points, it may appear correct in the processing but be difficult to notice shifts in other locations. By establishing validation points, you can check positional and elevation consistency before using point cloud data in practice. When treated as public surveying or as client-specified survey deliverables, accuracy control in accordance with applicable procedures, manuals, and specifications is required.

When setting shooting conditions, ensuring sufficient overlap between images is fundamental. In point cloud generation, the three-dimensional shape is reconstructed from the same features appearing in multiple images. Therefore, if adjacent images have little overlap, the point cloud may have gaps and the alignment may become unstable. In areas where similar shapes repeat, such as solar panels, processing can easily be confused, so make sure that surfaces where features are easy to pick up—such as ground surfaces, walkways, slopes, and structures—are properly captured.

Pay attention to the camera orientation. Shooting straight down is well suited for capturing planar terrain, but slope faces, the sides of drainage channels, areas along fences, and the raised or vertical features around equipment can be difficult to capture. Combining oblique-angle shots as needed makes it easier to supplement side information. However, increasing oblique shots also increases the number of images to process and makes it more likely that unwanted background will be included, so use them according to your purpose.

The condition of grass and debris also affects point cloud quality. In areas where grass is overgrown, point clouds are more likely to capture the surfaces of the grass, making it difficult to confirm the actual ground surface. At solar power plants, the appearance of the ground surface can change significantly before and after mowing. When performing regular comparisons, it is important to record the site conditions at the time of imaging, because differences in grass condition alone can make the terrain appear to have changed.

Step 4 Capture the site with a drone

Once preparations are complete, drone shooting is carried out on site. On the day of shooting, we first perform a pre-flight inspection and check the aircraft, the transmitter, the battery, the recording media, the camera settings, and the status of position information acquisition. Even if a flight plan has been prepared in advance, adjustments may be necessary on site due to local wind, worker movements, vehicle traffic, or surrounding safety conditions. We prioritize obtaining the necessary images safely over flying strictly according to the plan.

While shooting, check that images are not excessively dark or blown out. Solar panels are prone to reflections, and under some conditions parts of them can become overly bright and blown out. Because point cloud creation requires extracting features from the images, having many images that are extremely difficult to see will affect the processing results. Since reviewing all images after shooting and then re-shooting is time-consuming, it is reassuring to proceed while checking sample images on site.

When photographing an entire power plant, proceed while confirming that the flight route has not deviated and that no areas have been missed. The edges of the site, along fences, the tops and bottoms of slopes, and areas around drainage facilities are places that easily fall outside the flight coverage. In particular, when point cloud data will be used in explanatory or comparative materials, missing critical areas may be impossible to supplement later. Conduct supplementary shooting as needed and add images of priority areas.

When performing oblique photography, choose an angle that captures the subject adequately while maintaining a safe distance. If you want to check the terrain around slopes, steps, drainage channels, or mounting racks, shooting straight down can make it difficult to understand the shapes. Shooting from an oblique angle makes it easier to capture three-dimensional forms and the condition of the sides. However, if there are many reflections on panel surfaces or repetitive patterns, image processing can produce incorrect results, so ensure the ground surface and fixed objects are properly captured.

Don’t forget to keep a record of the capture. Recording the flight date and time, weather, wind conditions, capture area, status of control point placement, grass condition, any anomalies noticed on site, and areas that were re-shot — even briefly — will help with later processing and result verification. Looking at the point cloud data alone won’t reveal the conditions at the time of capture. Especially for regular inspections or post-disaster comparisons, differences in capture conditions affect how results appear, so the records serve as important context for judgment.

Step 5 Generate point cloud data from images

We generate point cloud data from the captured images. In this process, we proceed with tasks such as importing images, aligning them, setting control points, generating a high-density point cloud, cleaning up unwanted points, and verifying coordinates. The processing itself is generally performed using dedicated image-processing software, but the important thing is not pressing buttons—it's advancing while confirming that the processing results meet the intended purpose.

The first thing to do is organize the images. Blurry images, extremely dark images, images with severe overexposure, or images that are not relevant can affect processing quality. By separating images by shooting order or by area and excluding unnecessary images before processing, it becomes easier to verify the results. In large power plants, it can be easier to find problem areas by dividing the images into sections and checking them rather than processing all images at once.

Image alignment estimates the positional relationships among multiple captured images. If this fails, the point cloud can appear duplicated, rows of panels may bend unnaturally, or the terrain may look wavy. In solar power plants, because similar-shaped panels are repeated, image processing can mistakenly match corresponding points. Whether features such as pathways, ground surfaces, drainage channels, fences, and buildings are properly captured affects the stability.

When using control points, correctly mark the control points on the images and correlate them with the coordinates measured on site. If the specified positions of the control points are offset, the position and elevation of the entire point cloud will be affected. If the markers appear too small in the images or are difficult to see due to shadows, the placement error can increase. It is important to photograph the control points before processing so they can be identified clearly.

After generating the high-density point cloud, check for unnecessary points and noise. This may include points floating in mid-air, points distorted by reflections, moving vehicles or people, variations caused by swaying grass, and unnatural points around panel surfaces. Rather than using the point cloud data as-is, organize it as needed and prepare it to suit the inspection purposes. However, because excessive editing can alter the true on-site conditions, it is prudent to record which areas were edited and how.

Choose the output format of point cloud data according to its intended use. The appropriate form varies by purpose: formats suited for checking in 3D display, formats that are easy to use for creating plan views and cross-sections, and lightweight formats that are easy to share with stakeholders. For power plant maintenance, organizing orthoimages, elevation data, and cross-section reference materials together with the point cloud itself makes the information easier to understand for people beyond on-site personnel.

Step 6 Confirm the results and apply them to on-site operations

When point cloud data has been generated, verify the results and use them in field operations. The important thing is not to judge solely on whether the point cloud looks visually clean. Check consistency with control points and verification points, any gaps in the target coverage, point cloud density, noise, how the ground surface appears, and the reproduction state of important areas. Point cloud data for solar power plants is one of the materials used to assess the condition of the plant, and should be handled in combination with on-site photographs and inspection records as necessary.

First, confirm that the target area has been captured as planned. Check that the areas included in the scope for the purpose—site boundaries, panel rows, maintenance walkways, drainage channels, slopes, fence lines, access roads, etc.—are not missing. In large power plants, edges and places with elevation changes are easily omitted, so check not only the overall view but also zoom in on key locations. If there are omissions, determine whether supplementary photography is needed or whether the existing data can meet the purpose.

Next, verify the validity of heights and shapes. Check whether parts of the point cloud are unnaturally floating, whether the ground appears wavy, or whether panel rows or access ways are shaped differently from reality. If there are reference points or validation points, judge the reliability of the results by looking at their differences. If the purpose is to confirm terrain changes at the power plant, it is especially important to ensure that the positions line up with past data. If coordinate or elevation assumptions are not consistent, places that have not actually changed can appear to have changed.

As a use case, the primary one is assessing current conditions. Because the entire large solar power plant can be viewed three-dimensionally, not only the personnel who visited the site but also managers and other stakeholders can more easily share the situation. Since panel layout, access paths, slopes, drainage routes, and surrounding terrain can be seen in a single dataset, it also helps when planning inspections and repairs.

Point cloud data are effective even for post-disaster inspections. After heavy rain or typhoons, if slope failures, sediment inflow, drainage channel blockages, deformation of maintenance access paths, or scouring of the ground are suspected, drone surveying can inspect wide areas and preserve the results as point cloud data. An advantage is that you can assess the situation from above before attempting to walk into areas that are difficult to access. However, it is important not to determine hazard levels based solely on point clouds; when necessary, follow up with on-site inspections and expert judgment.

In periodic inspections, you can repeatedly photograph the same location to check trends in change. Because vegetation condition, imaging conditions, and the handling of reference points differ each time and make comparison difficult, work procedures should be standardized as much as possible. By regularly accumulating point cloud data, it becomes easier to find not only abnormalities at a given time but also trends of subsidence or erosion that progress over time. This is a useful concept for the long-term maintenance of a power plant.

Common pitfalls to watch out for when creating point cloud data

When creating point cloud data for solar power plants, there are several common pitfalls. The most frequent is capturing images with unclear objectives. If you assume that simply photographing the whole site is sufficient, you may end up with insufficient resolution in required areas or, conversely, collect far too many unnecessary images. Point cloud data only delivers value when what will be checked after its creation has been decided in advance. Before capturing, it is important to clearly define the locations you want to inspect and the information you will use to make decisions.

Next, there is a lack of adequate shooting conditions. If images have little overlap, heavy shadows, strong reflections, many blurred images, or mainly capture areas with few distinguishing features, the point cloud can be incomplete or distorted. In solar power plants, where identical panels are repeated, image processing can become unstable. It is necessary to properly capture the ground surface, walkways, and surrounding structures to secure images that make it easy for the processing to determine positional relationships.

Insufficient reference points or an uneven distribution of them are also matters to be cautious about. If the 3D data is only for visual appearance, problems may not surface easily, but when checking distances or heights, or comparing with past data, coordinate consistency becomes important. Even if reference points are installed, if they are hard to see in images, difficult to designate, or concentrated in part of the area, the reliability of the results can decrease. The more point cloud data are used for explanation or decision-making, the more important the concept of reference and validation points becomes.

Differences in vegetation and seasonal conditions are also often overlooked. The appearance of point clouds representing the ground surface within a power plant changes significantly between immediately after mowing and when the grass has grown. If the grass surface is treated as the ground, there is a risk of misjudging subsidence or heaving. When making periodic comparisons, record the condition of the grass and the time of imaging, and avoid judging based solely on simple height differences.

Furthermore, point cloud data often go unused after being created. Point clouds are specialized data, and not all stakeholders can work with them as-is. For use on site, it is effective to organize them into materials that annotate the locations to be checked, cross-sections, comparison diagrams, and simple explanatory images. Only after being processed into a form that anyone can understand do point cloud data become easy to use for inspections, reporting, consultations, and repair planning.

Finally, it is also important not to draw too many conclusions from point cloud data alone. Point clouds created by drone surveys are an effective means of efficiently understanding wide areas, but they cannot directly verify subsurface conditions, the interiors of structures, or electrical faults. Use them as material for grasping changes in terrain and surface appearance, and in practice combine them with ground inspections, surveying, structural checks, and equipment inspections as needed.

Summary

To create point cloud data of a solar power plant by drone surveying, it is important to start by clarifying the objectives. Determine the area of interest and the items to be checked, carry out flight planning and safety checks, and after setting control points and imaging conditions, capture the site. Afterwards, generate point cloud data through image processing, verify the validity of the deliverables, and then use them for inspections and maintenance, post-disaster checks, pre- and post-construction comparisons, and explanations to stakeholders.

Drone surveying is an effective means of efficiently assessing large solar farms. However, the process of generating point clouds from captured images involves many considerations, such as image overlap, reflections, shadows, vegetation condition, control points, verification points, and checking the processing results. It is important to confirm not only that the visuals are easy to understand, but also that the data can be used for on-site decision-making.

To produce point cloud data that is practical for field use, it is essential to standardize procedures to match the power plant’s management objectives rather than carrying out each task on an ad-hoc basis. By preparing data with anticipated use cases in mind—regular inspections, post-disaster checks, repair planning, report preparation, and so on—drone surveying becomes not merely a photography task but an information foundation that supports the operations and maintenance of solar power plants.

When considering the creation of point cloud data and the use of drone surveys at a solar power plant, it is important to plan comprehensively, including on-site safety checks, the information you want to obtain, and how the results will be used. Rather than assuming specific equipment or methods, organize the plant’s scale, terrain, management objectives, required accuracy, and how you will explain things to stakeholders, and develop operational procedures suited to the site conditions; this will lead to continued maintenance and management.