josh_drone

Radio Tower Modelling

Recommended Posts

Hi Guys,

Does anybody model radio towers or masts?  If so, what is the optimal setup for getting images to build a Photogrammetry model?  I currently use a Phantom 4, Inspire 1 Pro and PIX4D.  Is it worth upgrading to a Matrice or similar, and getting a mirrorless camera (ie. Sony A6000) to take photos?  Thanks for any advice in advance.

Share this post


Link to post
Share on other sites

That’s a decision where your customer might have the best advice.  Do you have any examples of the models you’ve produced you can share?  I’m just curious are you shooting obliques or NADIR?  What ground control application are you using to create your mission?

if your customer is happy with the 3D meshes your providing now, I’m not sure why your thinking of changing your workflow?  I think the camera on the P4 is about 20MP and the Sony a6500 is 24MP, not a significant improvement.  The P4 has a michanical shutter, although I don’t think it’s a true global shutter, it’s much better than the rolling shutter on the a6500, however you can put a much better lens on the Sony.  

You should consider the Sony a7Rlll, that’s a full frame 43MP camera that is a significant upgrade over the P4.  You might want to consider upgrading photogrammetry software to something like ContextCapture from Bentley Systems.  

Another thing to consider is that there’s been an increase in people scanning cell towers with LiDAR.  I don’t know how anyone could afford to scan towers with LiDAR and be profitable.  But it means that if American Tower or Crown Castle expects models with that sort sort of accuracy then if your using Photogrammetry you’ll need to provide models with much better detail and also provide video inspections to try and compete.  

Sorry, I don’t think there’s a simple answer to your question.  It’s going to become a tough market to be competitive.  

 

 

 

Share this post


Link to post
Share on other sites

Hi mate,

I work for a telco company in NZ and the skies the limit really (pardon the pun).  I would like to generate the most accurate models possible in order to take dimensions off the model for checking steel / antenna / cable sizes.  I currently use pretty basic home made chequer board GCP's but will also start using an EMLID Reach RS to get more accurate GPS locations using RTK.  If there is a better workflow using photogrammetry I'm open to suggestions, however if LiDAR or 3D scanning is going to be more accurate then I could head down that path as well.

Cheers.

Capture.JPG

Share this post


Link to post
Share on other sites

We scan a lot of utilities.  Here's a scan of a substation.  I've been able to share this because PG&E used this model in a presentation at the AEC conference in Singapore.  This is about a six acre substation and it wasn't out more than 2cm in any direction.

http://realitymodeling-pw.bentley.com/a3D/Construct Timelapse/models/4-12/App/?scene=http://realitymodeling-pw.bentley.com/a3D/PGE/Las Aguilas/Rough Draft/scene/production_1.3mx#%2F

This is a lower resolution online version.  You need to let it cache but in the high res version you could read the serial numbers on much of the equipment.

We don't use ground control points or RTK.  What's the purpose of RTK?  Its to overcome the inaccuracy of GPS.  You can accomplish that much better by lessoning the latency and doing the math differently.  Everything in aviation is time and distance,  GPS is only one of the variables that effects accuracy and its not even the most important one.  

Probably the best solution will be blending LiDAR and Photogrammetry but for that to happen the relationship between the sensors and the object being scanned will have to be much more accurate.      

Share this post


Link to post
Share on other sites

Do you think the Bentley software is better than PIX4D?  From what I've read, PIX4D is supposed to be the premium one.  Another issue I have is colour noise up the tower as its not always possible to cut out all the sky when the structures are over 100m tall.  Does Bentley have a colour filtering component?

Share this post


Link to post
Share on other sites

You should probably check out Bentley Systems -- https://www.bentley.com/en 

I'm not saying that Pix4D isn't good software, it is the leader for using photogrammetry in agriculture, but Bentley, AutoDesk, Trimble and others also develop the software civil engineers use to design the bridges, buildings, and cell towers for example in the first place.  So it really depends on your customer.  Almost all of our customers use Bentley's MicroStation and design software and for the few that dont we use context capture anyway.  Those customers seem to like the quality of the 3D mesh.  That's a pretty picture thing.

Most of the time when we've used Pix4D its been for NADIR scans, almost all of our scans are obliques and Pix4D doesn't handle those nearly as well as ContextCapture,  However having said that all photogrammetry software has a tough time when you go above the horizon.  

Regarding color filtering, I don't know.  I'm the boss, although I have used ContextCapture I'm not an expert with it.  I just go from the reactions of our customers.  If their happy, I'm happy...

Share this post


Link to post
Share on other sites

Talon Aerolytics provides a service to it's major tower customers here in the US that includes creating 3D point cloud surveys to load into the owner's database. They host you for a 3 day training session ($185) so you can become adept at capturing the images they require. Once you meet their training certification you then are enlisted as a qualified pilot in their subcontractor's pilot network and available for callouts based on your location and travel radius preferences. One tower inspection with their program takes about 30-45 minutes for completion, 1.5 P4P batteries. Their only camera requirement is 20M pixel camera. When taking images for photogrametric purposes a global shutter is often better than a rolling shutter. POI images with a full 360 degree orbit of the top of the tower kept in the middle of the frame is one flight requirement. So if anyone is planning to use I2 then X4S is the better camera choice and something like Litchi will help with the POI orbit requirement. Oh the going rate for a single tower mission is $100. Many of their pilots are using P4P or I1 with several extra batteries or remote battery recharge capability.

The bottom line is that once each of the towers has it's own digitized photogrammetric data set created, the need for UAS pilots to fly the towers is greatly reduced to post storm damage assessment, periodic maintenance interval inspection and or new installs of towers or new equipment added to a tower.

So if you plan on making money flying towers get good quickly and hit it hard fast as the competition is ramping up and the need to fly them will diminish over time.

That is my 2 cents worth of conventional wisdom, worth almost as much as what you paid for it.

Share this post


Link to post
Share on other sites

@Drone Driver this is the problem with this industry.  Anyone can create a website and claim anything  like—“We choose the best available sUAS drone and sensor technologies”

Does anyone actually believe they provide the best drones for $100!?  They may choose the best “available” but that doesn’t mean it will be adequate.  I can pretty much say that it won’t be.  

There are so many variables that need to be accounted for to create accurate surveys that there’s no way this type of service will add the required value to either the customer or the operator.  

 

Share this post


Link to post
Share on other sites

Hey Chuck,

What outputs are available with contextcapture?  The problem with PIX is you really need the software to navigate the model and review the photos.  Uploading a model to sketchfab or PIX may look nice but its not that useful for clients.  The model can be exported to AutoCAD but the file is very large and slow in the software.  Another option is export the point cloud to ArcGIS but once again the client would need to have this software.

Share this post


Link to post
Share on other sites

@Av8Chuck

Reread my post, no where is it stated that a $100 drone is provided or used. The $100 information is the total gross payrate to the pilot per each tower inspection.

The only comments relative to equipment are 20M pixel camera, P4P, I1 and I2, we all know those machines cost more than $100 and no one sells a 20M pixel camera plus gimbal for a drone for $100.

Sorry if I gave anyone the impression that Talon was providing, selling or using $100 drones as part of these jobs. Just search out Talon Aerolytics, near West Point, Ga and you will see that they focus on creating the 3D point cloud as a deliverable to their clients. That is where they make their big money. As such is their business model, they will drive the payscale to pilots for data acquisition flights into the ground.

At $100/tower flight mission, the pilot would be doing good to net $60 as ALL insurance, logistics cost and or lodging/food are the responsibility of the pilot. So if you were able to get this work from their pilot provider subcontractor, you must have a multi-tower opportunity for towers which are close to each other and fly 5-6 per day.

My intended pearl of wisdom is that if all the tower owner/operators are able to get 3D point clouds for all of their towers, then the need to fly towers will be greatly dimminished. Once you acquire that level of data on each tower, it only requires an update at a much reduced interval.

Share this post


Link to post
Share on other sites

More info with this reply from Av8Chuck posted in my PM inbox - Thanks for your time clarifying and adding the details of what you are providing this service for your clients. I agree that Talon is attempting to drive the data acquisition price point into the ground and make it not worthwhile to get involved. However time will tell as to just how many towers they get inspected and inserted into their database.

The text below is from Av8Chuck:

I understood your post and I see what Talon is claiming to do.  We have done scans for both American Tower and Crown Castle, two of the largest cell tower companies in the US and they have paid us, depending on location about $1000 per tower.

The reason for the discrepancy in the price between what Talon is claiming and what we're charging is because what they are offering doesn't work.  This is not an IT problem its a problem of being able to meet the data requirements of the customer.  If you can't then there's no value in the data.  There are too many variables that don't get accounted for when using a P4 to provide the level of accuracy required.

Don't get me wrong, if they want to try that business model I wish them well.  But we don't go after that business because we don't even want to do it for $1000 per tower.  By the time you add in all the insurance, cost of equipment, travel expenses, etc, even at a $1000 per tower the margins are really thin.

Would you travel 25 miles to scan a tower for $100?  Plus the overwhelming percentage of phantom owners would have no experience at doing this.  There's no way Talon is going to get enough qualified operators with good enough equipment no matter what they claim on their website.

Besides, American Tower and Crown Castle want LiDAR which costs about $60K for one that meets their requirements.  There's quite a few companies trying to get the price down for Flash LiDAR but even then there's no way anyone could afford to provide scanning services for $100.

Here's a video of one of the towers we did:  

http://iplayerhd.com/player/82da4bbd-3097-4c0a-ad99-a94680b70139

Share this post


Link to post
Share on other sites

Hi mate,

Just wondering what outputs do you give the client?  Video, photos and a point cloud in AutoCAD?  Does contextcapture have any other output types?

Cheers,

Share this post


Link to post
Share on other sites

Hi @josh_drone we only supplied a 3D mesh from ContextCapture, the prime contractor supplied a written report, about 15 pages, a lot of information about the equipment on the tower.  i have no idea what it was or what it meant but the report was comprehensive and impressive.

I think it’s a bit nieve for people to think if they have a Phantom or Inspire and Pix4d they can do cell tower inspections.  There’s some serious engineering going on these towers. 

Share this post


Link to post
Share on other sites

Taking internal photos of the tower, either with a drone or with a panoramic camera, ie. fly up the middle.

When I run a model in PIX, it picks up colours from framing or sky on the other side of the tower, so the model isn't very clean.  It would be nice to get photos of the backside of antennas and internal framing as well.

Share this post


Link to post
Share on other sites
On 31/01/2018 at 3:55 AM, Av8Chuck said:

That’s a decision where your customer might have the best advice.  Do you have any examples of the models you’ve produced you can share?  I’m just curious are you shooting obliques or NADIR?  What ground control application are you using to create your mission?

if your customer is happy with the 3D meshes your providing now, I’m not sure why your thinking of changing your workflow?  I think the camera on the P4 is about 20MP and the Sony a6500 is 24MP, not a significant improvement.  The P4 has a michanical shutter, although I don’t think it’s a true global shutter, it’s much better than the rolling shutter on the a6500, however you can put a much better lens on the Sony.  

You should consider the Sony a7Rlll, that’s a full frame 43MP camera that is a significant upgrade over the P4.  You might want to consider upgrading photogrammetry software to something like ContextCapture from Bentley Systems.  

Another thing to consider is that there’s been an increase in people scanning cell towers with LiDAR.  I don’t know how anyone could afford to scan towers with LiDAR and be profitable.  But it means that if American Tower or Crown Castle expects models with that sort sort of accuracy then if your using Photogrammetry you’ll need to provide models with much better detail and also provide video inspections to try and compete.  

Sorry, I don’t think there’s a simple answer to your question.  It’s going to become a tough market to be competitive.  

 

 

 

Hi again,

Looking at getting a new drone setup for inspections and photogrammetry, probably the Matrice 600 and a mirrorless camera.  Is the Sony alpha 7 the only one you recommend?  Just wondering if there's a cheaper alternative..

Cheers.

Share this post


Link to post
Share on other sites

Interesting discussion. I manage hundreds of cell/radio towers for one of the big tower companies and I’ve been doing it for 17 years. I got a drone last year and started flying my towers. 

My employer is getting into the 3D modeling aspect but I can’t see the usefulness of it, yet. So far it seems to me to just kind of be artsy fartsy droning. 

As the final user of a 3d model, I can’t think of any circumstance on the towers and tenants I manage where 3d imaging is as practical as what I can accomplish as I always have. Flying the tower with regular video mode , and knowing what I know and my industry, all I need is my regular drone video. 

What am I missing?

If the reasoning is good enough, I could elect to participate in the 3D imaging program. I just don’t see it though. 

Share this post


Link to post
Share on other sites

Hi @Russ Putnam, I'm in a similar position but in New Zealand and Australia.  I currently inspect using a Phantom 4 and a Inspire Pro but want to upgrade to get better images, video and photogrammetry data sets.  Do you have any recommendations?  Thanks in advance.

Cheers, J

Share this post


Link to post
Share on other sites
12 minutes ago, josh_drone said:

Hi @Russ Putnam, I'm in a similar position but in New Zealand and Australia.  I currently inspect using a Phantom 4 and a Inspire Pro but want to upgrade to get better images, video and photogrammetry data sets.  Do you have any recommendations?  Thanks in advance.

Cheers, J

Hi Josh, not really. I’m the ultimate rookie, kinda. I just started last Fall and I’ve flown 47 towers since then. I know all my towers well enough I’m just TRYING to find reasons to fly. 

Mostly, it’s just to get line of sight video of equipment that’s difficult to see from the ground, for auditing purposes. 

I have found some deficiencies , saved some tower climb costs and a few other things. Because I’m so busy, I try to limits my flights to 9 minutes, which is 4gb , which is one single video file. 

What I try to do is refocus the camera often throughout the flight to maintain crisp quality. Sometimes when o forget to refocus then my video is blurry later when I’m watching it. 

If I wasn’t so busy with such a large area and so many towers, the 3d imaging would be fun, but thus far, I don’t see it holding a candle to a crisp video. 

Ive got a lot to learn though, still. 

 

Best, 

 

Russ

Share this post


Link to post
Share on other sites

There's a bit of a difference between 3D surveys and 2D video inspections.  3D models, generally produced with Photogrammetry. LiDAR or both become three dimensional relational databases where a lot of meta data can be appended to the model so the owner of the tower can simply click on any part of the model and get the born-on-date, installation, who installed it, warranty information etc.   

Video inspections are real time that can provide some high resolution imagery and help determine what if any additional inspection is required.

I'm not sure which is better, depends on how people want to use them.  Although I doubt Crown Castle or American Tower are willing to pay enough to make it a profitable business.  

Share this post


Link to post
Share on other sites

Back to profitability statement again. This is my concern as for some customers the stills and or video product is satisfactory. But for others the 3D point cloud is a requirement.

So my question then becomes who pays for the significant difference in time and fees for post processing software and or computer time to create the 3D point cloud?

Pix4D has the option of either local processing if you have enough computer to accomplish or cloud processing. Either is time consuming and adds a cost premium to the data acquisition equation. Can a pilot who gets paid for collecting the images also charge the added costs for providing a 3D point cloud with photogrammetric precision to the customer?

Again I suspect that Av8Chuck would correctly state in a reply - that is a customers choice or preference.

Then I would ask if his company has a Bentely Systems (or Pix4D) license plus the expertise to digitally edit and or align these images for premier accuracy output, then process the images into a deliverable 3D point cloud?

The bottom line to me is that I wish to know what is the potential for adding profit to a set of images which you collect as the drone pilot if you produce the 3D point cloud?

I have some initial budget estimates put together for a fast deskside computer and costs of software applications as well as cloud computing license costs, it is difficult for me to make this comparison.

Share this post


Link to post
Share on other sites

Hello, I coincidentally came onto this blog. I am not a pilot but I used to run Pix4D in North America until the acquisition by Parrot. This is my perspective:

. It is very difficult to fly towers and capture the images well that a good photogrammetric model is created - it takes flight skills *and* understanding of how the photogrammetric process work so images are taken with the software in mind

. photogrammetry for structurally small elements like lattices or cables is difficult for the software, and there are many variables that impact the 3D reconstruction such as what the ground looks like. Image-based point clouds are noisy for cell towers, which means lots of manual cleaning and tie points. LiDAR point clouds are less noisy, but of course the acquisition cost (machine, sensor, insurance, etc) is much higher and currently not commonly done this way.

. 3D tower models should command a high price, as they require lots of skills and experience. The end-user (eg tower company) understands that but needs those prices to come down to make their cell tower inspection at scale work well. Everything has to come down: flight control software to be more automated (difficult with all types of towers, guide wires, etc); software license price, compute time and compute hw, conversing of point clouds to vectors, decrease manual software work, etc...

. as pilots, to digitize a tower, you need to own, be trained, and have hours of practice on the software. Those skills are much different than those of flying. 

. sure, you can deliver a point cloud file and mesh of a tower, but what value really does it have to the end user? End users want reports. They want to know what they need to replace or fix, to know what real estate is left on that tower. In addition to the answers (not data), they might also want the tools to consume the model generated. A dataset of 600 images and a 600 million points .las file has little usability to them. 

. if you, as a pilot (most pilots are quite astute technically (I know some of you are engineers and surveyors!)), decide to provide modeled radio/cell towers, work with the end-users to understand how they use the data, what software they have, the reporting format needed, what metrics they expect from you, how you can self-perform QC before you deliver, etc...

My 2 cents!

Antoine

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



  • Welcome to UCCF.

    The UAV Coach Community Forum is actively moderated by the UAV Coach team and offered to help serve those in the UAV industry. Use this space to meet and greet, to ask and answer questions, to share what you're working on, etc. Have fun, play nice, and fly safe! :)

  • Forum Statistics

    • Total Topics
      3,078
    • Total Posts
      15,334