We are starting to look at using UAV's with total stations for mapping crash scenes and was interested if anyone out there had some good ideas for workflows. The UAV is likely to be a DJI Phantom 4 RTF and the software is Pix 4d Mapper. I am not sure what camera is going to be used.
Pat
Workflow for UAV and total station for crash scenes
-
- I have made 20-30 posts
- Posts: 29
- Joined: Thu Nov 26, 2009 12:31 pm
- 14
- Full Name: Pat MCDONALD
- Company Details: QPS
- Company Position Title: Crash Investigator
- Country: Australia
- Linkedin Profile: Yes
- Location: Gold Coast, Queensland
- Contact:
-
- V.I.P Member
- Posts: 846
- Joined: Tue Feb 26, 2013 9:32 pm
- 11
- Full Name: Massimo De Marchi
- Company Details: Massimo De Marchi
- Company Position Title: freelancer
- Country: Italy
- Skype Name: massimo_de_marchi
- Has thanked: 15 times
- Been thanked: 53 times
- Contact:
Re: Workflow for UAV and total station for crash scenes
HI,
phantom 4 has its integrated camera, so no options there.
Use a tablet or phone base flight planner (Altizure, Pix4d, drones made easy, drone deploy, ..), and test it.
In general you can fly with any app and process in a different one.
Check the pix4d help, there is an excel sheet to estimate GSD based on flight altitude.
Use a 80% overlap all the time.
My personal opinions about apps:
Altizure: easy, fast to setup, can fly 5 missions (zenith and 4 with tilted camera) to have a nice reconstruction. Only rectangular shapes (you can rotate the rectangle).
Pix4D: nice freeflight option
drone deploy: arbitrary shape, you must log in.
drones made easy: only for apple products, a lot of options, with a paid option you can also upload a terrain model
To find the correct worklow you have to do some tests, but:
I would start flying a full altizure mission and tagging a handful of points (5-10) with total station.
The only parameter missing is flying height. For a crash scene ad a baseline I would fly at 40-50 meters from ground.
The reconstructed area is inside the path with the camera at 90 degrees.
Massimo
EDIT: 40-50 meters means a GSD of about 2 cm
phantom 4 has its integrated camera, so no options there.
Use a tablet or phone base flight planner (Altizure, Pix4d, drones made easy, drone deploy, ..), and test it.
In general you can fly with any app and process in a different one.
Check the pix4d help, there is an excel sheet to estimate GSD based on flight altitude.
Use a 80% overlap all the time.
My personal opinions about apps:
Altizure: easy, fast to setup, can fly 5 missions (zenith and 4 with tilted camera) to have a nice reconstruction. Only rectangular shapes (you can rotate the rectangle).
Pix4D: nice freeflight option
drone deploy: arbitrary shape, you must log in.
drones made easy: only for apple products, a lot of options, with a paid option you can also upload a terrain model
To find the correct worklow you have to do some tests, but:
I would start flying a full altizure mission and tagging a handful of points (5-10) with total station.
The only parameter missing is flying height. For a crash scene ad a baseline I would fly at 40-50 meters from ground.
The reconstructed area is inside the path with the camera at 90 degrees.
Massimo
EDIT: 40-50 meters means a GSD of about 2 cm
ing. Massimo De Marchi - +39 347 32 17 049 - www.studiodemarchi.net
- vislaw
- V.I.P Member
- Posts: 125
- Joined: Thu Feb 17, 2011 1:42 pm
- 13
- Full Name: Mark Johnson
- Company Details: Visual Law Group
- Company Position Title: Principal and Chief Creative Officer
- Country: USA
- Linkedin Profile: Yes
- Location: San Jose, California
- Has thanked: 1 time
- Been thanked: 6 times
- Contact:
Re: Workflow for UAV and total station for crash scenes
We have been mapping collision scenes using drones and Pix4D for a few years now. Although we have a total station, we rarely incorporate it in our workflow because we get sufficient accuracy for our purposes without it. We are not concerned with integrating our scenes geospatially. We just need to be certain the generated scene is properly scaled and that measurements made from it are accurate to the degree our clients require.
Since much of our work ends up in court we partnered with one of the engineering firms we often work with to do our own validation study. The firm cut an assortment of PVC tubes to random lengths and manually scattered them over an open field. We were not given the measured lengths beforehand so that our derived measurements would be truly blind.
We first flew a DJI Inspire drone at low altitude (50') and flew a mapping grid. Finally, we used a Topcon total station to locate the position of the PVC tubes as well as 4 checkerboard targets as control points. We then processed the drone imagery without any control point correction and projected the total station data points on top of the drone orthographic image. Without any correction the spatial error was about 2%, which is pretty decent considering the relatively low tolerance of drone gps systems. Of course, accuracy is highly dependent on the altitude flown and the number of overlapping images for a given area of interest. Pix4D has a good discussion of these issues https://support.pix4d.com/hc/en-us/arti ... #gsc.tab=0.
As long as your mapping grid is reasonably uniform your spatial errors will be consistent across your scene. Thus, you can get acceptable accuracy by simply measuring a known distance at the scene and conforming your derived map to that landmark. We typically grab a couple measurements from roadway lane markers and use those to conform our final map. Doing so will give you a generated map accurate within an inch for your primary area of interest if your references are carefully obtained.
The important thing we learned is that a good drone grid will produce a map that has a very good relative scale so that rectifying it to match a single known distance goes a long way toward giving you a usable diagram. Adding additional ground control points, of course, will dial it in very precisely.
Certainly we understand that there are occasions when we need very tight tolerances and in those instances we will bring the total station (or, more likely, our Faro scanner) out to a scene. But in many instances our clients are content to use roller-tape measured skids and lane widths so that the drone matches or exceeds the police report diagram without using any correction.
Since much of our work ends up in court we partnered with one of the engineering firms we often work with to do our own validation study. The firm cut an assortment of PVC tubes to random lengths and manually scattered them over an open field. We were not given the measured lengths beforehand so that our derived measurements would be truly blind.
We first flew a DJI Inspire drone at low altitude (50') and flew a mapping grid. Finally, we used a Topcon total station to locate the position of the PVC tubes as well as 4 checkerboard targets as control points. We then processed the drone imagery without any control point correction and projected the total station data points on top of the drone orthographic image. Without any correction the spatial error was about 2%, which is pretty decent considering the relatively low tolerance of drone gps systems. Of course, accuracy is highly dependent on the altitude flown and the number of overlapping images for a given area of interest. Pix4D has a good discussion of these issues https://support.pix4d.com/hc/en-us/arti ... #gsc.tab=0.
As long as your mapping grid is reasonably uniform your spatial errors will be consistent across your scene. Thus, you can get acceptable accuracy by simply measuring a known distance at the scene and conforming your derived map to that landmark. We typically grab a couple measurements from roadway lane markers and use those to conform our final map. Doing so will give you a generated map accurate within an inch for your primary area of interest if your references are carefully obtained.
The important thing we learned is that a good drone grid will produce a map that has a very good relative scale so that rectifying it to match a single known distance goes a long way toward giving you a usable diagram. Adding additional ground control points, of course, will dial it in very precisely.
Certainly we understand that there are occasions when we need very tight tolerances and in those instances we will bring the total station (or, more likely, our Faro scanner) out to a scene. But in many instances our clients are content to use roller-tape measured skids and lane widths so that the drone matches or exceeds the police report diagram without using any correction.
-
- I have made 20-30 posts
- Posts: 29
- Joined: Thu Nov 26, 2009 12:31 pm
- 14
- Full Name: Pat MCDONALD
- Company Details: QPS
- Company Position Title: Crash Investigator
- Country: Australia
- Linkedin Profile: Yes
- Location: Gold Coast, Queensland
- Contact:
Re: Workflow for UAV and total station for crash scenes
Thanks very much for the information very helpful
Pat
Pat