iPad Pro (Wi-Fi + Cellular)

PIX4Dcatch PIX4Dcatch Questions/Troubleshooting
1

Hi Pix4D Support,
Does the Pix4Dcatch app benefit from the iPad Pro with Wi-Fi + Cellular model for remote outdoor captures as it has a built-in GPS/GNSS module?

iPad Pro (Gen 2) with LiDAR

  • iPad Pro (Wi-Fi + Cellular) has built-in GPS/GNSS
  • iPad Pro (Wi-Fi) has a digital compass (iBeacon microlocation). Therefore, it does NOT have GPS /GNSS module.

I understand that iBeacon microlocation positional accuracy is based on the number of available Wi-Fi routers it’s able to locate as a function of distance. This is a wonderful strategy for mapping indoor environments where Wi-Fi is available, but how does this strategy support outdoor remote data captures. On construction sites, Wi-Fi may not be available. Even if there is WiFi on a remote site, it would only serve as a single iBeacon. I would assume that a dual or tri-iBeacon would be a minimum requirement to provide reasonable positional accuracy.

I had tested the Pix4Dcatch app on an iPad Pro (Gen2) without cellular, on a couple remote job sites which consisted of 500-1000 images. None of these model stitch properly. Recently, I was re-evaluating the Pix4Dcatch app, when I turned-off Wi-Fi and a message in the app said “turn on Wi-Fi to improve location accuracy”. This is an important topic that Pix4D support should define in detail, as it pertains to the iPad Pro (Gen2) model (Wi-Fi vs. cellular) and it’s data positional accuracy in various environments.

Those of us on the UAV-side understand that GPS and image overlap are two very important functions in the image stitching algorithm. I believe it’s safe to assume that this also remains true for ground terrestrial data capture. Unless, the small/cheap cellular GPS/GNSS module is not accurate enough for ground capture. If so, I assume the app relies more heavily on LiDAR and IMU data for image stitching. Therefore, the stitching algorithms for the iPad Pro is different because the iPad data captures at a smaller-scale (ie. mm-cm range). In comparison to a drone that typically collects data over a greater distance (ie. meters), GPS data is a primary factor in the stitching algorithm.

I understand that Pix4D may not want to discourage iPad Pro (Wi-Fi) users against deploying/testing the new app, as Wi-Fi only iPad models makeup 90% of sales. However, for those of us who want to ensure a quality stitch, it’s best to understand the benefits of the cellular model now for early adoption. Users would be frustrated to learn that the iPad Pro (Wi-Fi only) device is presenting significant positional inaccuracy in outdoor environments, which could have otherwise been avoided at the Point-of-Purchase. We would like to be made aware if the cellular model iPad Pro is recommended now or even in the future (algorithm updates), for remote site captures where Wi-Fi is non-existent.

From a users perspective, learning that a 3D model did not stitch properly is not a gamble worth taking if we had initially bought the cellular iPad Pro model. It’s only $150 difference for an iPad Pro model with Wi-Fi + cellular. Knowing that the Pix4Dcatch app uses the cellular GPS/GNSS data is important to the users. We need to know what environment (indoor vs. outdoor) the iPad Pro model is more suitable when capturing data.

I understand people want to use the Pix4Dcatch app to merge aerial and terrestrial models (myself included), but are having a great level of difficulty doing so. I wonder if this is due to the lack of accuracy iBeacon microlocation data presents. Pix4D support, have you successfully merged the two models together (sample link)? Can this be compiled on the cloud or must the projects be merged on the desktop version of Pix4Dmapper?

Kind Regards,

2

Hello,

It is true that that the iPad Pro without Cellular does not have a GPS/GNSS module, this can lead to a less accurate position of the project.

However, Pix4Dcatch only uses the geolocation to find the position of the project, and the location is sampled throughout the duration of the capture to try to get the most accurate result. This position is then used in another calculation to give the final positions of the images in the projects. Pix4Dcatch uses other techniques to keep track of it’s own position in the area you are capturing, so that all images are located relative to each other with very high accuracy, even if the overall placement of the project is off, even by a large margin.

In the case of a project that has poor positional accuracy, the whole project can be transformed to the correct location manually. Or GCPs can be used during capture, as you might with drones, to place the project exactly.

If a project failed to work at all, then another issue is likely, we would be glad to look into it, but to techniques used by Pix4Dcatch to accurately position images within the project, it doesn’t actually matter to it if it has geolocation or not, iPads with and without cellular should still give the same results in terms of the reconstruction, even if the overall location of the project is less accurate.

We are of course looking into getting the best accuracy we can for catch and it is something we are working on.

As to stitching multiple projects, perhaps this article can help : https://support.pix4d.com/hc/en-us/articles/202558529-How-to-merge-projects

or maybe others here can help.

Kind Regards,

George Brown

3

Hi George,
When you say “this can lead to a less accurate position of the project”.
Does this suggest that using an iPad Pro (Wi-Fi) on a remote worksite (no Wi-Fi available) the tablet will not produce an initial position for the project? If so, is the model create in a “black box” that needs to be geolocated in Pix4Dmapper? A benefit to the iPad Pro (Wi-Fi + cellular) is that it will get an accurate position in these outdoor work environments (worry free!). Overall, it sounds like the iPad Pro (Wi-Fi + cellular) is more accurate model for those looking to deploying Pix4Dcatch outdoors (ie. worksites) where Wi-Fi is scarce or non-existent. However, can the user hotspot the location from their phone at the beginning of capture to over-ride the iBeacon microlocation issue as stated above?

Thank you for outlining the method for the Pix4Dcatch app. I had suspected the algorithm relied heavily on the LiDAR and IMU to stitch the images. Is there a maximum number of photos the algorithm works best within? Is it possible that 500-1000 images were too many? I haven’t seen much about the best capture strategy using the iPad Pro from Pix4D. I would assume that when capturing in indoor areas, the iPad can be angled from +90 to -90deg, whereas in outdoor areas, the iPad should remain 0 - -90deg. I’m assuming it’s best to limit the amount of sky/sun (aka. illumination conditions and time of day) in the outdoor imagery (Unless facing a building where +20deg may stitch.

I’m hoping my initial question will help others select an iPad Pro to ensure they get the results they desire. I also hope that they can better understand the workflow upon initial deployment. This app is very impactful and I love what your team at Pix4D has created. Look forward to hearing from you soon.
Kind Regards,

1 Like
4

Hello,
In the case that no geolocation is acquired, the result will be at null island (0,0).
The iPad pros were the first generation of LiDAR enabled devices by Apple and we see the utilisation of that technology as a major step forward in the abilities of consumer level devices in our field, the new iPhone 12 pro also offer this capability, and we’re also working on an android version that also have equivalent sensors
We have, for the moment, placed a hard limit on the number of photos taken to 1500, tbh, for performance reasons, we have tested further than that and got results. It’s a very difficult question to answer…
I’m sure many here are aware that there is a trade off of many factors in the number of images captured, the overlap, the subject matter that can affect the process in many ways, I’m very confident that Pix4Dcatch can achieve the results our customers require, and we look to ways to ensure we get the best results, we are very keen to make the app work in an easy to understand manner and customer feedback is incredibly valuable, if you would be prepared to share with us the results of any failed projects, we will be glad to look into it

5

Will catch work indoors or is GPS required? If so I would like to see a webinar on the process, please

6

Hi Brad
Catch should work indoors. Unfortunately, we don’t have a webinar organized for now.

Kind regards
Amritha

7

Thank you for this topic and discussion! I wish I found this thread before I purchased a few 2024 iPad Pro’s 11-in (WiFi only) to run Pix4Dcatch for outdoor use. I used it to do an outdoor scan of a courtyard where WiFi is available. The result was not very good as a tremendous amount of drift both horizontally and vertically occurred. I used Pix4Dmatic to process the 3D model mesh that I opened in Rhino. The way I walked the site may have contributed to the issues with overabundant overlapping of images, but I have not experienced this level of drifting with my iPhone 15 Pro Max. I wish there were more videos demonstrating best practices with scanning.

8

First of all, thank you for this entire discussion. I tried using an iPad Pro (WiFi-only) in a remote area without WiFi signal and captured approximately 2,300 images using a side-by-side scanning pattern. However, I noticed that the initial positions are all showing up at Null Island (0,0).

My issue arises when I try processing the data without georeferencing—it produces a good orthomosaic. But as soon as I georeference with around 5 control points, I encounter data gaps. Even though the position aligns with my required coordinate system, the output product doesn’t turn out well.

When you say:
“In the case of a project that has poor positional accuracy, the whole project can be transformed to the correct location manually. Or GCPs can be used during capture”

Is there any way to fix the issue I’m facing?
I have posted more detail about this issue (with quality report and processing logs) in following group as well: Pix4D catch data processing noises in DSM