Processing options
      Back to home
      1. Pix4D Documentation
      2. PIX4Dmatic
      3. Processing options

      Dense point cloud - PIX4Dmatic

       

      Pix4Dmatic raycloud icon

      Calibrate

      Pix4Dmatic point cloud icon

      Point cloud

      PIX4Dmatic_mesh_icon.png

      Mesh

      Pix4Dmatic dsm icon

      DSM

      Pix4Dmatic orthomosaic icon

      Orthomosaic
      PIX4Dmatic_point_cloud_icon_small.png Dense point cloud PIX4Dmatic_depth_point_cloud_small.png Depth point cloud PIX4Dmatic_fusion_depth_dense_small.png Depth & dense fusion

      Dense point cloud

      The Dense point cloud step generates a dense point cloud from images.

      Access:
      • Click Process processing_options.png.
      • On the Menu bar, click Process > Densify...

      It allows users to change the following processing options:

      Dense_Point_cloud

      Algorithm

      This option allows switching between the densification algorithms:

      • Hardware accelerated (Default): The densification will be computed using a GPU-capable algorithm. This option can contribute to reduced processing time.
      • Standard: The densification will be computed using the standard algorithm.
      Note: The Hardware accelerated algorithm is automatically selected if a compatible GPU is installed (any NVIDIA GPU that supports OpenGL 4.1 or higher). Check here on How to verify the supported OpenGL versions of the graphics card.
      Note: The Hardware accelerated algorithm was tested on Windows 10 and we compared it to the Standard algorithm. The tests showed that the densification processing time is on average 52% faster using the HW accelerated algorithm compared it to the Standard algorithm. We also report a slight increase in the number of densified points.

      Image scale

      Image Scale defines the scale of the images at which additional 3D points are computed.

      • 1/1 (Original image size, Slow): The original image size is used to compute additional 3D points. This option require four times more RAM and processing time than when processing with the default value 1/2, and usually it does not significantly improve results.
      • 1/2 (Half image size, Default): Half size images are used to compute additional 3D points. It is the recommended image scale.
      • 1/4 (Quarter image size, Fast): Quarter size images are used to compute additional 3D points. Less points are computed than with the 1/2 image scale and the processing time is reduced. This option can generate more points when processing vegetation areas.
      • 1/8 (Eighth image size, Tolerant): Eighth size images are used to compute additional 3D points. Less points are computed than with the 1/2 or 1/4 image scale and the processing time is further reduced. This option can generate more points when processing vegetation areas.

      Multiscale

      Multiscale (default): When this option is activated, additional 3D points are computed on multiple image scales, starting with the selected Image scale and going to the 1/8 image scale.

      Example: If 1/2 image scale is selected, additional points are computed on images with 1/2, 1/4, and 1/8 image scale.
      Tip: We recommend to enable the Multiscale option when:
      • Computing additional points in vegetation areas as well as keeping details in areas without vegetation.
      • Filling holes in places with low texture, for example, uniform walls.
      The Multiscale option may generate additional noise, we recommend to disable the Multiscale option when:
      • The edges of buildings contain noise.
      • When reconstructing thin elements, for example, the structure of the cell tower.

      Density

      This parameter defines the density of the point cloud:

      • Optimal (Default): A 3D point is computed for every 8th pixel of the original image. This is the recommended point cloud density.
      • High: A 3D point is computed for every 2nd pixel of the original image. It results in an oversampled point cloud that requires up to 4 times more processing time and RAM than Optimal density. Typically, this point cloud option does not significantly improve the results compared to Optimal density.
      • Low: A 3D point is computed for every 32nd pixel of the original image. The final point cloud is computed up to 4 times faster and uses up to 4 times less RAM than Optimal density.
      Note: The point Density has an impact on the processing time and the number of 3D points generated.

      Minumum matches

      Minimum Number of Matches (2-6, 3 default) represents the minimum number of valid re-projections of this 3D point to the images. For example, when using the Minimum Number of Matches 3 (default), each 3D point has to be correctly re-projected in at least 3 images.

      Tip: We recommend using the Minimum Number of Matches:
      • 2: For projects with a small overlap. Note that this option produces a point cloud with more noise and artifacts.
      • 4, 5, or 6: To reduce the noise and improve the quality of the point cloud. Note that this option might compute less 3D points in the final point cloud.

      Noise filter

      Noise filter processing option provides a cleaner point clouds for datasets with oblique images.

      It filters out points that are generated far away from images for example, features that are visible on images but are far on the horizon will not be reconstructed.

      Example: When the Noise filter is enabled for a cell tower project, points that are far away from the images do not get reconstructed.

      PIX4Dmatic_noise_filter_enabled

      PIX4Dmatic_noise_filter_disabled

      Sky filter

      Sky filter processing option removes points in a dense point cloud associated with sky.

      The feature is especially useful for datasets that contain sky in images, for example:

      • Terrestrial datasets acquired with PIX4Dcatch.
      • Inspection datasets acquired with PIX4Dscan.

      PIX4Dmatic sky filter

      Important: The Sky filter processing option significantly increases the processing time. We recommend using it only for datasets that contain a lot of sky in images.

      Mask-aware

      Masks can enhance the quality of dense point clouds and meshes by removing selected elements from the scene, effectively hiding them from 3D. Specify the layers to be used in the Dense point cloud and Mesh steps under the Mask-aware setting.

      PIX4Dmatic_Mask_aware

      Exports

      From the Exports section, the following point clouds can be exported after processing the Densify and Mesh step:

      • Dense point cloud (available after processing Densify step)
      • Point cloud from mesh .laz (available after processing Mesh step)

      PIX4Dmatic export point clouds

       

      Dense point cloud

      To export the point cloud automatically during the Densify step:

      • In Exports, select the Dense point cloud .laz option.
      • (Optional) Click Pix4Dmatic folder and specify the desired path and filename.
      Note: The Dense point cloud .laz option is disabled by default. It is possible to export the point cloud without re-running Densify step:
      • On the Menu tab, click File > Export dense point cloud .laz...
      • Specify the desired path and filename.
      • Click Save.

       

      Pointcloud from mesh .laz

      PIX4Dmatic allows the generation and export of point clouds from the mesh. The default Density (pts/m3) or a customized one can be selected while exporting the new point cloud from the mesh. The expected point cloud size and the total number of points are also displayed while exporting the point clouds. This way, a point cloud is generated even for areas that are not directly reconstructed and holes no longer exist in the reconstructed point cloud. The point cloud generated from the mesh is in .laz format.

      Note: This point cloud is expected to have a much bigger size than the point cloud files that are generated during the Densify step.
      PIX4Dmatic export pointcloud from Mesh