Case Study UAV Use on a Crash Scene Versus Total Station Sergeant Daniel Marek Nevada Highway Patrol

Transcription

1 Case Study UAV Use on a Crash Scene Versus Total Station Sergeant Daniel Marek Nevada Highway Patrol

2 How comfortable is your couch??

3 Case selected for: First use of a UAV by Nevada law enforcement under a FAA Certificate of Authorization (COA) Traditional mapping with Total Station was completed. Non-prosecution fatal crash Mistakes in UAV use to learn from

4 Summary On Saturday, September 17, 2016, at approximately 1110 hours, a 1993 Harley-Davidson motorcycle (V-1) was traveling eastbound on SR168, west of mile marker 6. The rider (D-1) attempted to negotiate a left curve at speeds too fast for conditions. V-1 ran off the south edge of the roadway. D-1 maintained control of V-1 in a straight line to the southeast. The front of V-1 ran over a large rock causing V-1 to overturn prior to a cliff. D-1 was ejected from V-1 as both travelled over the cliff s edge. D-1 came to rest in a prone position, facing southeast, at the bottom of the cliff, south of SR168. V-1 came to rest on its left side facing south, east of D-1.

5 Leica TS12 Total Station

6

7

8

9 Approximately three Hours of work on scene and two hours in the office on the computer. Drawn with Map Scenes 2010 software.

10 The next wave of crash scene documentation is to use a UAV to take various over head images and create a 2D orthomosaic map image of the scene. The software used in this presentation is trial versions of the various software. I have had no training on any of the software and processed all of the images as they were captured from the UAV without any alterations to improve the processing. This is not an endorsement of any software or manufacture product.

11 Manual flight over scene conducted with Yuneec Typhoon H. This was the first law enforcement use of a UAV in Nevada.

12 This is the first photo ever taken by a UAV for law enforcement purposes in Nevada.

13

14

15

16

17 Orthomosaic map of crash scene process in Pix4D. Orthomosaic is an aerial photograph or image geometrically corrected ("orthorectified") such that the scale is uniform: the photo has the same lack of distortion as a map

18 Orthomosaic map of crash scene processed in DroneDeploy.

19 Total station On scene: 3 hours Office: 2 hours Total: 5 hours Typhoon H Flight: 18 minutes Pix4D 58 minutes Total 1 hour 16 minutes DroneDeploy 34 minutes Total 52minutes Note: processing time can occur during the drive back to the office.

20 Pix4D measurement of the furrow: 101 yards. Note the trash on the right edge due to some oblique photos not processing correct.

21 Horizontal distance of the 19.6 yards of fall.

22 3D view rotated image showing the 20.8 yard diagonal fall distance.

23 Pix4D comparison of the generated orthomosaic to satellite map.

24 Sample reconstruction using Pix4D Given fall distance horizontal 19.6 yards or 58.8 feet diagonal 20.8 yards or 62.4 feet Solved vertical (Pythagorean Theorem) as 6.96 yards or 20.9 feet Will use an assumed takeoff angle to be level. Will also assume that the rock struck at point of departure did not influence the fall. Will also assume a level grade of the furrow of 101 yards or 303 feet, and a drag factor of.48 for the soft dirt and rolling friction of motorcycle in gear

25 Fall equation d= 58.8, g=32.2, G=0, h=-20.9 = 51.6 f/s or 35 mph at take off

26 Initial velocity equation Ve = 51.6ft/s, a=-32.2(.48), d= 303 feet v v v v v i i i i i = = = = v e ad (2* 32.2*.48*303) = fps = 74mph Initial velocity at roadway edge 109 fps or 74 mph. This is not the true speed in this crash. Grade of the furrow could add to final speed as would the down slope at takeoff point.

27 DroneDeploy comparison Biggest things I immediately noticed were a surface length measurement, the slope and elevation profile.

28 Fall (blue line) Fall horizontal distance of feet. Height of feet (Pix4D numbers were 58.8 feet and 20.9 feet)

29 Furrow final downhill (red) Surface length of feet with a grade of -8.34%

30 Furrow uphill (orange) Surface length of feet with a grade of 0.9%

31 Furrow initial downhill (yellow) Surface length of feet with a grade of -1.46%

32 Sample reconstruction using DroneDeploy Given fall measurements feet distance, height feet Takeoff angle -8.34% Three initial velocity equations for the different slopes a drag factor of.48 for the soft dirt and rolling friction of motorcycle in gear feet with a grade of -8.34% feet with a grade of 0.9% feet with a grade of -1.46%

33 The math results from DroneDeploy Start of fall fps or 37.9 mph Start of final downhill fps or mph Start of uphill fps or mph Leaving the road surface fps or mph

34 Speed Comparisons Pix4D DroneDeploy Speed at start of fall 35.1 mph 37.9 mph Speed at start of final downhill n/a 53.1 mph Speed at start of uphill n/a 60.5 mph Speed at leaving the roadway surface 74.7 mph 74.8mph

35 What could have prevented the crash? More level surface: 2 2 Ve Vi d = 2a d = 2*32.2* d = d = 86 feet

36 Slower speed: 9 mph slower at the point the motorcycle left the roadway would have allowed the rider to stop at the cliff edge. v where, v v v v v i i i i i = v e 2 e 2ad = 0 = 0 2ad = 0 (2*32.2*.48*303) = = fps = 65.97mph

37 Errors and things to consider Manual flight does not result in proper image capture for proper processing. (need 70-85% overlap of images) Some trash in Pix4D image due to the errors in photo capture and import. All of this done without any training on Pix4D or DroneDeploy. Trial and error learning.

38 Positive takeaway With automated flight and image capture many errors can be eliminated. (now capable) Cloud processing can allow investigator to upload from scene and have map done by the time the investigator gets to the office. Time saving on scene.

39 Final thoughts

40 Don t let this be your couch.

41 Thank to the following