Blog | Importing Very Large Photogrammetric Data Sets into Virtual CRASH
By: Marc Edgcombe, Traffic Reconstruction and Transportation Compliance
Updated: July 17, 2019
The combination of Virtual CRASH and photogrammetric software such as Pix4D has become a powerful tool in the crash reconstruction arena when used with small Unmanned Aerial Systems (sUAS), a laser scanner, or a combination of the two. For the purposes of this post, we will discuss working with large data sets collected with aerial platforms and the use of Pix4D in combination with Virtual CRASH 4.
You can learn more about importing and scaling an orthomosaic image, and aligning it with point cloud data, in this Blog Post.
When importing orthomosaics into Virtual CRASH, you might occasionally encounter an issue in which Virtual CRASH doesn’t appear to respond to the file you’re attempting to import. This is usually caused by either (1) an internal error with the orthomosaic image itself or (2) attempting to import an extremely large orthomosaic file (typically over 1 GB). The fix in both cases is relatively simple:
1. First, attempt to import the orthomosaic file into a 3rd party graphics editing program such as Gimp (which is a free app). At times, the process used by Pix4D to export the orthomosaic can induce certain errors into the file that can cause it to fail to open in several different programs, including Virtual CRASH. Importing and subsequently re-exporting from Gimp using the JPEG settings will fix these file errors and maintain the majority of the original quality. If an orthomosaic of any size is not importing into Virtual CRASH, try this method first. The screenshot below shows the menu option for export (File>Export As>*select file name and location and press ‘Export’, the ‘type’ menu shown will then appear*). Note this will not solve the file size issue, only the file internal error issue.
2. Second, the issue may be the size of your orthomosaic as it’s exported from Pix4D. You have two options at this point, but first, make sure your settings are set to 1 cm/pixel upon export of the orthomosaic. This can be done in the following Pix4D processing menu. Note the “custom” value may refuse to be manually changed by the user until the “automatic” setting has been selected and manipulated up or down. At this point the user can re-select and assign the “custom” value (this seems to be a feature in Pix4D). See the following screen shot:
After this action has been taken, the orthomosaic can be processed and imported into Virtual CRASH upon completion of processing. This also virtually negates the requirement to scale the orthomosaic upon importation into Virtual CRASH due to the fact that Virtual CRASH currently assumes 1 cm/pixel. Best practice is to verify the scale, although it should already be very close.
3. At this point, if your orthomosaic is too large to import into Virtual CRASH you can import orthomosaic ‘tiles’, or portions of the orthomosaic for key areas of interest and combine them together. These tiles are located in the Pix4D folder, usually located at *file name*>3_dsm_ortho>2_mosaic>tiles. Select what tiles you need and then import them into Virtual CRASH. Keep in mind if you’re working with extremely large orthomosaic you may only be able to import a certain number of tiles before the application’s memory allotment is exceeded. At this point, if you attempt to import additional tiles they will show up as white squares with no image content.
An example of these tiles can be seen in the following screen shot:
4. If you do not wish to import individual tiles, the next solution is to go back to your Pix4D project and edit the Processing Area. This can be done after “1. Initial Processing” or after “2. Point Cloud and Mesh” but always before “3. DSM, Orthomosaic and Index” is processed.
A. Using it after Step 1 will cut your processing time for the Step 2 and Step 3 processing but will limit the point cloud processing to the area you selected for processing.
B. Using it after Step 2 will retain the entire point cloud for processing if you decide you want it later but will limit the Orthomosaic processing to the area you selected. This is a good option if you want the entire point cloud but only a portion of the point cloud to be processed into the orthomosaic. This is the option I used to reduce the size of the orthomosaic and import it successfully into Virtual CRASH.
5. If, after the size of the orthomosaic is reduced in Pix4D using the processing area and is exported, but it still will not import into Virtual CRASH, one or two issues may be taking place:
A. The orthomosaic is still too big
B. The orthomosaic contains errors and needs to be imported into Gimp to be re-exported.
If either of these is the case, see the previous steps to remedy the issue. Screen shots of the processing area can be seen below:
First, select the processing area icon. This will allow you to create a polygon in the workspace to outline the area of interest.
After you have completed your polygon, right click to complete the area. Pix4D will then remove anything outside the perimeter of your area. You can then adjust the minimum and maximum altitude of the Processing Area on the right side of the screen. This is useful if you have artifacts below the surface or tall objects you would like excluded from the point cloud. At this point, you can export your point cloud and it will only include what is within the Processing Area. You can also initialize Step 3 in order to process the orthomosaic with only the data contained within the processing area. Remember to set the resolution to reflect the expected 1 cm/pixel. If you require the entire point cloud to be exported, you can export the point cloud prior to creating the Processing Area or you can delete the processing area and Pix4D will re-populate the point cloud.
Finally, remember, as the creator and designer of your simulation or animation environment, you must decide what image size makes sense to use for your scene. For example, while one can generate an amazingly detailed orthomosaic image of a road surface using photographs from a low-flying drone, where even small cracks in the asphalt are clearly visible, it may not make much sense to use such a highly detailed (and typically very large) image if the camera you use to render your scene is placed 100 feet above the road surface. In such a case, those small surface details will not be visible in the final animation.
Byron Center, MI