Getting started

Best practices and capabilities for PIX4Dcloud

Before you start your project, please carefully review our guide on the best practices and capabilities of PIX4Dcloud.

This document outlines essential prerequisites for optimal results and details the software's processing capabilities and effectiveness with specific features. Also, please be sure to familiarize yourself with the terms governing credit refunds or allowances for images/projects to understand eligibility criteria. Your diligence in understanding these guidelines is crucial for a smooth and successful use of PIX4Dcloud for your project.

Best practices with PIX4Dcloud

Plan the mission and capture images according to the project area or object you want to reconstruct.


General case

The recommended overlap for most cases is at least 75% frontal overlap (with respect to the flight direction) and at least 60% side overlap (between flying tracks). It is recommended to take the images with a regular grid pattern (Figure 1). The camera should be maintained as much as possible at a constant height over the terrain/object to ensure the desired GSD.

flight_plan_2.png
Figure 1. Ideal Image Acquisition Plan - General case.

Forest and dense vegetation

Trees and dense vegetation often appear very differently between overlapping images due to their complex geometry (thousands of branches and leaves). Therefore, extracting common characteristic points (keypoints) between the images is more difficult. To achieve good results, it is recommended to use a grid image acquisition plan like the one described in the General Case section by applying the following changes:

  • Increase the overlap between images to at least 85% frontal and side overlap.
  • Increase the flight height: At higher altitudes, there is less perspective distortion (therefore causing fewer appearance problems), and the dense vegetation has better visual properties. In other words, detecting visual similarities between overlapping images in such areas is easier. The flight height, image pixel resolution, and focal length determine the images' ground sampling distance (spatial resolution). Best results are obtained with a GSD higher than 10cm/pixel.

Flat terrain with agricultural fields

In cases where the terrain is flat with homogeneous visual content, such as agriculture fields, it isn't easy to extract common characteristic points (keypoints) between the images. To achieve good results, it is recommended to use a grid image acquisition plan like the one described in the General Case section by applying the following changes:

  • Increase the overlap between images to at least 80% frontal and side overlap.
  • Fly higher. In most cases, flying higher improves the results.
  • Have accurate image geolocation and use the Ag multispectral template.

Building reconstruction

Reconstructing 3D buildings requires a specific image acquisition plan (Figure 2):

  • Orient the pitch of your camera so that the majority of the image is filled with the object you want to reconstruct, and objects you do not want to reconstruct comprise the minority of the image. No single camera pitch angle can be applied to all missions.
  • Fly a second and third time around the building, increasing the flight height and decreasing the camera angle with each round.
Note: For more information about oblique imagery: Vertical vs oblique imagery.
  • It is recommended to take one image every 5-10 degrees to ensure enough overlap, depending on the size of the object and its distance to it.
Note:
  • The flight height should be increased at most twice between the flights, as different heights lead to different spatial resolutions.
  • PIX4Dcloud generates a high-quality point cloud for oblique images of buildings. However, no orthomosaic is generated when the selected template is 3D Models.
building_reconstruction.png
Figure 2. Ideal Image Acquisition Plan - Building.

Special cases

This section presents some hints for terrain that is difficult to map, such as terrains with snow, sand, lakes, etc.

Snow and sand have little visual content due to large, uniform areas. Therefore:

  • Use a high overlap: At least 85% frontal overlap and at least 70% side overlap.
  • Set the exposure settings accordingly to get as much contrast as possible in each image.

Water surfaces have almost no visual content due to large, uniform areas. Sun reflection on the water and waves cannot be used for visual matching.

  • Oceans are impossible to reconstruct.
  • Each image must include land or some other stationary object to reconstruct other water surfaces, such as rivers or lakes. Flying higher may help to capture more land features.

Corridor mapping

Mapping a corridor such as a railway, road, or river requires at least two flight lines (Figure 3). GCPs are not required but are recommended to improve the accuracy of the reconstruction. For more information about the number and distribution of GCPs in corridor mapping: Number and distribution of ground control points (GCPs) in corridor mapping.

For a dual track, it is recommended to use at least 85% frontal overlap and at least 60% side overlap.

It is possible to use nadir images or oblique images. For flat terrain, it is recommended to use nadir images.

dual_track.png
Figure 3. Dual track image acquisition plan for corridor mapping.

If a dual-track image acquisition plan is not possible, a single-track image acquisition plan can be used if (Figure 4):

  • Overlap is high enough: At least 85% frontal overlap.
  • Ground control points (GCPs) are defined along the flight line in a zigzag pattern.
single_track.png
Figure 4. Single-track flight NOT RECOMMENDED.

Large Vertical Object reconstruction

The 3D reconstruction of objects that are tall and slender requires a specific image acquisition plan (Figure 8):

  • Fly close to the structure.
  • Fly several times around the structure at several heights.
  • Images should be taken with high overlap: 90% overlap between images taken at the same height and 60% overlap between images taken at different heights.
  • Everything in the image frame must be in focus, including objects in the background that are outside the project area.
  • Having image geolocation is recommended.

power_reconstruction_2.png
Power Tower
power_reconstruction.png
Power Tower reconstructed in the rayCloud
Figure 8. Image Acquisition Plan - Power tower.

Note: For more information on how to map and measure pole and tower structures: How to map and measure Pole and Tower Structures.

Thermal

For a better reconstruction of the captured scene in a thermal project, some recommendations should be followed during the image acquisition:

  • Have very high overlap: 90% front and side image overlap.
  • The images have been taken at a resolution of at least 640x480.
  • The images do not suffer from motion blur. An increased flight speed may cause a blurred image.
Warning: PIX4Dcloud/PIX4Dcloud Advanced only supports JPG grayscale images. Note that the results may have scale problems.

Capture images with an appropriate ground sample distance.

This article explains what the Ground Sampling Distance (GSD) is and how to calculate it.

The Ground Sampling Distance (GSD) is the distance between two consecutive pixel centers measured on the ground. The bigger the value of the image GSD, the lower the spatial resolution of the image and the less visible details.

The GSD is calculated based on:

  • The flight height / the distance from the terrain or object (H).
  • The camera specifications:
    • Image width (ImW).
    • Sensor width (SW).
    • Focal length (F)

Understanding_the_relationship_between_sensor_height_and_GSD.png

Information: 

It is important to decide on the GSD value before starting the image acquisition in order to adjust the flight height and the camera specifications to the project requirements. For example, when a detailed reconstruction of the area is needed, we recommend flying closer (low GSD) to the object of interest. On the other hand, when covering large areas that do not need very detailed results, flying higher (high GSD) can greatly reduce the acquisition time and batteries needed as well as to reduce the processing time.

Example: 
  • A GSD of 5 cm means that one pixel in the image represents linearly 5 cm on the ground (5*5 = 25 square centimeters).
  • A GSD of 30 cm means that one pixel in the image represents linearly 30 cm on the ground (30*30 = 900 square centimeters)

Pix4D_ground_sample_distance_GSD_5cm.jpg

Pix4D_ground_sample_distance_GSD_30cm.jpg

Note: Even when flying at a constant height, the images of a project may not have the same GSD. This is due to terrain elevation differences and changes in the camera angle while shooting. Since the orthomosaic is created using the 3D point cloud and the camera positions, an average GSD will be computed and used.

Capture images that are detailed, sharp, and contain enough contrast.

Consider the following aspects to capture images that are detailed, sharp, and contain enough contrast.

Camera Settings

The shutter speed, aperture, and ISO should be set on automatic. If images are blurry or noisy, it is recommended to manually set these parameters.

There is a tradeoff between the shutter speed, the aperture, and the ISO sensitivity. For processing, the images should be sharp and have the least amount of noise. Such images can be obtained when the scene is well illuminated (scattered clouds should be avoided), and the camera parameters are well adjusted. If the scene is not sufficiently illuminated, images will be noisier and less sharp, thus lowering the accuracy of the results.

  • As a rule of thumb, the shutter speed should be fixed, the ISO needs to be set at a low value that does not produce noisy images, and the aperture should be set to automatic to adjust for varying levels of brightness in the scene. If the tradeoff is not correct, overexposed or underexposed images may be obtained.

    The shutter speed should be fixed and set to a medium speed (as an indication: between 1/300 second and 1/800 second), but fast enough to not produce blurry images. If more than 5% of the images are subject to a directional blur, it is a good indication that the shutter speed should be slightly increased.

    The ISO should be set as low as possible (minimum 100). High ISO settings generally introduce noise into images and drastically reduce the quality of the results.

    The aperture minimum and maximum values depend on the lens. High aperture is translated into low numbers, for example, f2.7 (which will capture a lot of light). If both the shutter speed and ISO are adjusted, it is better to leave the aperture (f) on automatic.
  • The electronic and mechanical stabilization should be disabled as it interferes with Pix4Dmapper's algorithms.
  • The recommended focus mode is Manual Focus on Infinity. This mode of focusing should always give focused images for aerial projects. For terrestrial projects, this mode will probably lead to out of focus results, if a long focal length is used.

Problems with images due to wrong camera parameters or inadequate equipment that interfere with the processing:

Blur due to slow shutter speed.
Noise due to high ISO sensitivity.
Overexposed or underexposed (wrong aperture and/or shutter speed).
Distortions due to electronic or mechanical image stabilization.
Distortions due to the rolling shutter.

Select the processing options that are appropriate for the dataset.

PIX4Dcloud offers different processing options depending on the type of flight that the user would like to process. The recommended options are listed below:

Type of flight and dataset
Recommended processing options
Nadir RGB images
3D maps
Oblique RGB images (*)
3D model
Terrestrial RGB images
3D models
Multispectral images
AG multispectral
RGB images of agricultural fields
2D orthomosaic
Thermal images (**)
3D maps (low-quality results expected)

* Consider activating the Building reconstruction option to improve the quality of the outputs for building projects with homogeneous facades.

** Only JPG grayscale images are supported but not recommended. The results may exhibit radiometric errors, scale problems and calibration inaccuracies.

Information: PIX4Dcloud/PIX4Dcloud Advanced alone doesn't support processing RJPEG images. PIX4Dmapper can be used to process those images. See Processing thermal images using PIX4Dcloud/PIX4Dcloud Advanced and PIX4Dmapper.
Warning: Selecting the wrong processing option for your flight may lead to bad results.

For more information, see New dataset > Processing options > Templates.

(Optional) Improve the absolute accuracy by using GCPs or capturing images with an RTK drone.

The expected absolute accuracy of the final results (orthomosaic, DSM, point cloud, and DSM) when using PIX4Dcloud depends on the following:

Accuracy of the geolocation of the images

Type of drones
Expected accuracy on x,y
Expected accuracy on z
Standard drone
5 -10 m
10 m
RTK / PPK drone
1 cm +-2 ppm
2 cm +- 2 ppm

*ppm: parts per million ( 1mm per km)

Example: An RTK drone receives RTK corrections from a station 5 km from the project area.
  • Accuracy in X,Y → 1 cm + 2 * (5 mm) = 1.1 cm
  • Accuracy in Z → 2 cm + 2 * (5 mm) = 2.2 cm

For more information, Flying camera to surveying tool: RTK/PPK drone upgrades.

Use of GCPs

If ground control points (GCPs) are used, the absolute accuracy of the results will be comparable to the absolute accuracy of the GCPs, even with a standard drone. For example, if the accuracy of the GCPs is about 2-3 cm, we can expect that same degree of absolute accuracy in our final results.

More information: RTK vs PPK drones vs GCPs: which provides better results?

PIX4Dcloud Advanced includes the AutoGCP feature, which allows the use of GCPs to improve accuracy. Please see the requirements for this feature.

Capabilities of PIX4Dcloud

PIX4Dcloud can automatically detect GCPs when the following requirements are fulfilled.

Access: This feature is only available on PIX4Dcloud Advanced and when a 3D map or a 2D orthomosaic template is selected.

Image and GCP requirements

For the AutoGCPs detection algorithm to succeed, the following requirements regarding the images and the GCPs must be met. Additional information and best practices are available in Automatic detection of targets - AutoGCP algorithm.

Image requirements

  • Image quality has to be good (blurry or low-contrast images make the algorithm fail).
  • Images have to be geolocated (5-10 m accuracy is enough) in WGS84 (EPSG:4326).
  • Images have to be taken at nadir or slightly oblique angles.

GCP requirements

  • Squared, diagonal, and Aeropoints targets are supported.
  • The recommended size is a minimum of 20 times the average GSD of the project.
    Example: If the average GSD of the projects is 2.5 cm (1 inch), the recommended target size is 50 cm (20 inches).
  • Black and white targets are supported. Other colors may work in some circumstances.
  • GCPs must be placed at a distance of at least 10 m from each other.
  • GCPs must have a projected coordinate system.

targets.png

Warning: If the AutoGCPs detection algorithm fails, the processing will run with no GCPs.

PIX4Dcloud offers a default set of optimized options for processing.

This step allows the user to select a processing template for the project. Different processing templates are available:

  • 3D maps: Generates a point cloud, 3D textured mesh, orthomosaic, and DSM. It is ideal for aerial RGB images acquired using a grid flight plan with high overlap, mostly oriented towards the ground. Typical applications are quarries, cadasters, etc.
    Note that this template can be processed with different types of pipelines
    • PIX4Dmapper compatibility: This option processes the project with the PIX4Dmapper engine, which may present limitations in terms of processing power and the quality of the results. The Export to PIX4Dmapper option is available when the project is finished processing.
    • New pipeline: Currently, this pipeline is offered as the default processing engine in this template. This pipeline is only effective for nadir flights. Vertical coordinate systems and geoids are now supported compared to the old engine, with faster processing and high-quality results. For more information, New 3D maps processing pipeline.
  • 2D orthomosaic: Generates an orthomosaic. Ideal for RGB aerial images acquired using a grid flight plan with high overlap on flat terrain, for example, agricultural fields.
    The quality needed for the outputs can be specified between:
    • Quality: Good output quality with high accuracy. Processing times longer than the Rapid option are expected.
    • Rapid: Lower output quality. Faster for assessing the quality of the acquired dataset.
  • 3D models: Generates a 3D textured mesh and point cloud. Ideal for any RGB images with a high amount of overlap, such as images taken from the ground or oblique aerial images (free flight). Application examples are 3D models of buildings, objects, ground imagery, indoor imagery, inspection, etc.
    The quality needed for the outputs can be specified between:
    • Quality: Good output quality with high accuracy. Processing times longer than the Rapid option are expected.
        • Building reconstruction: Improve the quality of the outputs for building projects with homogeneous facades.
    • Rapid: Lower output quality. Faster for assessing the quality of the acquired dataset.
  • AG multispectral: Generates an NDVI map for precision agriculture. Ideal for aerial images from multispectral cameras with band-dedicated sensors, acquired with high overlap using a grid flight plan.
Important: The list of supported multispectral cameras and images can be found in How to process agriculture projects on PIX4Dcloud.

Processing_templates_PIX4Dcloud_v2.png

Once the processing template has been selected:

  1. (For PIX4Dcloud users) Select Create to finalize the dataset creation. The upload of the inputs starts.
    (For PIX4Dcloud Advanced users) Select Next and follow the instructions for step 4. GCPs automation.
Information: Sometimes, the default parameters may produce wrong results even if the flight is correct and the images are of sufficient quality.
Note: If the processing options must be changed, other software packages like PIX4Dmatic or PIX4Dmapper can be used instead.

PIX4Dcloud can generate and display reconstructions, but it cannot edit them.

PIX4Dcloud is a web-based software enabling the visualization of various outputs, including Orthophotos, Digital Surface Models, NDVI indexes, Point Clouds, and 3D meshes. It's important to note that these results cannot be edited within PIX4Dcloud.

We recommend using alternative products such as PIX4Dmapper, PIX4Dmatic, or PIX4Dsurvey for any necessary editing.

PIX4Dcloud can process datasets up to a certain size.

PIX4Dcloud can successfully process a certain number of images according to the characteristics of the project, the processing options, and the requested outputs.

Information: Users with a PIX4Dcloud or a PIX4Dcloud Advanced license are not allowed to process more than 4000 images per project. More information about cloud allowance at PIX4Dcloud - FAQ.

The variety and combinations of the parameters mentioned above can be many. This article is intended to provide some benchmarks and to show the results of some tests performed with significant camera resolutions, common project characteristics and default processing options.

Test results
 
Dataset A
Dataset B
Camera resolution
24 MP
42 MP
Camera used for testing
senseFly AeriaX
Sony Cyber-shot DSC-RX1RM2
Maximum number of images processed successfully
4000
2160
GSD
2.22 cm / 0.87 in
0.9 cm / 0.35 in
Overlap
75% - 75%
70% - 40%

Type of flight

Nadir, single grid
Nadir, single grid
Type of project
Industrial zone and agriculture area
Excavation and agriculture area
Processing template
3D maps, Quality¹
3D maps, Quality¹
Time to generate results
1 day 4 hours
19 hours

¹The 3D maps, Quality template includes the generation of the Point Cloud (.las), 3D Textured Mesh (.obj), Orthomosaic (GeoTIFF), and DSM (GeoTIFF). More information in New dataset - Processing template.

Warning: Higher quality outputs, additional outputs and output formats can affect the probability of success of the project processing.
Important: Note that the new 3D maps pipeline can process 3951 images of a 42.2 MP camera in a project. For more information, New 3D maps processing pipeline.

PIX4Dcloud can display reconstructions up to a certain size.

A dataset can consist of uploaded results only. In this case, no images are processed but the uploaded files are processed for their web-visualization and optimization.

The photogrammetric files that can be uploaded and displayed as results are the following:

  • Orthomosaic (.GeoTIFF),
  • DSM (.GeoTIFF),
  • Index (.GeoTIFF): NDVI map or other index maps
  • Point Cloud (.las, .laz),
  • 3D Textured Mesh (.obj).

There are some limits in terms of file size and formats for file uploads:

  • Orthomosaic: one single GeoTIFF file supported, 50 GB maximum 
  • DSM: one single GeoTIFF file supported, 50 GB maximum 
  • Index: one single GeoTIFF file supported, 50 GB maximum 
  • Point Cloud: one single LAS or LAZ file supported.
  • 3D Textured Mesh: one single OBJ+MTL+JPG set of files supported
    • OBJ: Maximum 1 GB.
    • JPG: Maximum 16k x 16k.

PIX4Dcloud uses default parameters for processing.

PIX4Dcloud is a user-friendly web-based software designed for simplicity, eliminating the need for expertise in photogrammetry. It uses a set of default parameters tailored for various flight scenarios. These parameters are not adjustable.

Warning: There are instances where correct flights and high-quality images may still yield inaccurate results due to the limitations of default settings. In such cases, alternative software solutions like PIX4Dmatic or PIX4Dmapper can be used for greater flexibility in adjusting processing options.
Information: If a dataset does not fulfill the best practices or requires more than the capabilities explained above, the allowance or credits it consumed will not be returned.