Camera work

I been reading the copter AP forums and looking at what people are using what it produces.  The Panasonic GH2 seems to be the ticket.  It's not a dslr but CMOS.  

These guys have been in the TV business and shot a nice video. 

They have business in Atlanta doing this.  It was my understanding this was not "leagal" but maybe they know something we don't.....

Pirate Eyes

We have tried to shoot some video from the copter and found piloting Line of Sight (LOS) is very difficult.  One needs to keep LOS to comply with current regulation for sUAV in US air space FPV guidelines  Currently most people use goggles with small LCD screens which precludes LOS flying.  A new product called Pirate Eyes has a small monocle that allows one to see video transmitted from the platform but also to have LOS on the copter.  Below is a review I found for RCGroups: 
Hello all,
I ordered a Pirate Eye in a desperate attempt to make our pilot's life a bit easier and to get better aerial shots in return. The main problem for the pilot was that he was at no point able to judge distances adequately and framing shots perfectly flying line of sight only. Our camera operator's ground station wasn't helping, as the pilot had to stand close by and look him over the shoulder every now an then to see the live-footage. 

So here's our two cents wrapped up in a little review:

We ordered the "hobby version", as the pro version wasn't available. Not a big deal, as the only difference is that the pro version has both the smokey and the clear glasses and comes in a case. 

In detail, our version came with the smokey glasses with the display and video adaptor attached, a nylon pouch, a cleaning cloth and a small goody bag containing an allen wrench,
distance pieces for the monocular, a neck strap and spare rubber rings for attaching the video cable to the glasses.

The overall quality of the set is great and IMHO justifies the price of around 400 Euros. If I look at the overall price of our aerial platform, this is peanuts... but as I said, that's just my personal opinion.

On to the important stuff: We only had the chance to try the Eye twice over the weekend for a couple of hours, but were initially pleased with the results. Setup is hassle free, the glasses fit perfectly and the display gave us crispy live-video out of a Canon 650D. Our initial worries that the tiny monitor
might fog-up as it frequently happens with closed systems like the Fatshark due to a lack of venting fortunately weren't confirmed. However, we have yet to try it under warm and humid conditions...

Flying was a breeze and we could see that the pilot's confidence got a boost. While maintaining a steady line of sight at most times looking straight ahead, he could easily 
look down at the display and frame shots precisely. For us that's a huge plus, as we are currently flying a no-pan-gimbal and all pans have to be done by the pilot...

I know this might sound a bit over-enthusiastic, but if a product caters perfectly to my needs, I'm more than happy to help promote it a bit. Especially when it's coming from a small niche company.

Just a final note to finish this up: The pirate eye is not made for FPV-flying. If you want to immerse yourself, buy a Fatshark or similar. If you are, however, into aerial video and photography, give the Eye a chance and try it out.

Hope that helps!

Cheers,
Danj

Caps and Cans

Since taking pictures we have been experiencing focusing problems with the Cannon T2i.  We are moving at 3 m/s at 60 m AGL and take pics every 1.5 seconds.  Other will need to chime in and fill in camera settings and transect information.  Somethings we have well focused pictures and other times not.  To eliminate copter vibration as a cause we balanced the props with a magnetic prop balancer:

 The light tip had drops of CA glue added to bring it into balance.  Next, we went to balancing the motor cans (without prop).  Right now we are just going by sound, feel and the copter ACC display in MKTools.  When we tried adding tape to motor #7 it made an awful screaming sound and the change in rpm goes in huge steps.  We soldered in another motor and we know the esc was bad.  This was our second esc to fail.  Each esc has a deformed area on the MOFSET closest to the inboard capacitor.  It looks as thought the cap has overheated and melted into the component.  The esc was ordered as well as new 355 uF/25V caps.  We will also inspect and replace all aftermarket caps and then silicon glue each down.

Update 10/2/12
I changed out esc #7 and #2.  Esc #2 had tp be change out because I tried to take out the capacitors and had to really heat up the board.  I did not want to risk a motor failure in flight so I changed it out.  I consider it $70 buck of insurance.  I also solder in one of the capacitor backwards.... all the anodes go on the outside of the board and are clearly marked.  I also used silicon to support the capacitors since this can be a source of vibration.  I made a table of motor wire and their connections to make assembly easier:

Motor       Color sequence (looking down on copter and going clockwise)
1               BPG
2               PGB
3               BPG
4               PGB
5               PBG
6               PGB
7               BPG
8               PGB
P-purple, G-grey, B-blue

I will continue balancing but this time I will use a laser and mirror like this:  http://www.youtube.com/watch?v=QnjflLDxQQU

Deep Zoom

One of the problems with the tremendous mosaics that we'll generate is actually exploring the images.  The technology for zooming, particularly on the web, is pretty tedious. Or at least was.  Looks like Deep Zoom might be something we can use to make our imagery widely available.  Check out a demo from Youtube.

Flat Mosaics

This tutorial should help us out for flat mosaics

Full field mosaic

mosaic attempt 2

Today we tried to use Hugin to create a photomosaic of some of the field images from 9/11/12. Hugin is freeware, available for OSX and Windows. Trying to use Windows didn't work very well, so the best option was to use my Mac.

Hugin was pretty easy to use - we used the Assistant which expedited the process. 

We're uncertain about the projection type still, but we may try this method next:

http://hugin.sourceforge.net/tutorials/Mosaic-mode/en.shtml

Mosaic Generation from Hugin

Today we tried to use Hugin to create a photomosaic of some of the field images from 9/11/12. Hugin is freeware, available for OSX and Windows. Trying to use Windows didn't work very well, so the best option was to use my Mac.

Hugin was pretty easy to use - we used the Assistant which expedited the process.

We're uncertain about the projection type still, but we may try this method next:

http://hugin.sourceforge.net/tutorials/Mosaic-mode/en.shtml

JR 9503 2.4 Setup

I did some reading ( https://sites.google.com/site/mkokto2/jr-x9503-setup ) on setting up the JR 9503 radio and changed these settings:

40% Exponential on elevator and ailerons
Reset throttle endpoints to -113 and +113

Added a Throttle curve:
L  0%
1  26.5%
2   46.5%
3   50%
4   53.5%
5   64%
H  100%

I flew these settings and altitude is easier to adjust.

mww

List of flights and file names

Here's a list of all the recent flights and files associated with them. For reference, we also flew on July 9th (the first flight and crash), July 13th (2nd crash), and July 20th (practice around Winston and the big field). There are only Iphone and IPad accounts of those images and videos though.

What is he doing out there?

I flew a quad for while but always lost orientation.  When I was flying a tricopter I could tell front and back much better. So I built a V-tail quad.  I have best of both worlds.  Excellent yaw authority but no mechanical linkages.  I also get the added stability of a quad but with better orientation from a distance.  Got it finish and programmed a multiwii board (Quadrino Zoom - 9 degrees of freedom, Acc, gyro, compass and barometer) a friend loaned me.  Did a late night flight with the LED's blazing in the neighborhood.  Heard my neighbors say, "What is he doing out there?"

Got the idea from here: http://www.rcgroups.com/forums/showthread.php?t=1622447

Used Arduino 1.0 sketch with V-tail and Quadrino Zoom commented.  Used the Multiwii 2.0 configuarator to setup board.  Powered board with a single bec from motor 3.  

Got to tune the PID's since I have some twitching.  Will update when/if I get it dialed in.

From the sky

Image from 58 meters of the team. 

The Art of Consistency

From our flights on 9-7-12 I copied out the Lat and Long values for two way point flights that only differed in their altitudes.  One was at 60 meters and the other at 30 m.  They look very close.

We also did a Point Of Interest (POI).  Here is it's flight path.  Not bad for a first attempt.

Start to Finish Image Processing Attempt 1

Here's a quick three step process to go from JPG and JPGw from Palentier to a projected geotiff:

1. Copy both the JPG and the JPGw files into a folder that you want to work with
2. Open up the program "OpenEV". Load each file, then export as a GeoTiff to the appropriate directory. Make sure the file extension is .tif   - do this for all images (one at a time now, but hopefully we can batch this process in the future). When you save the file, name the file without an extension, and then afterwards, while viewing it in windows, give it a .tif extension. For some reason, giving the file a .tif extension during the export will give you a B+W image.
3. Add the .tif file into Arc. This should automatically create an .aux file in the directory with projection information. This should translate and be able to be read by ENVI.

ArduCopter 2.7.3

There is a new release from ArduCopter that looks to have made major improvements.  It might be time to grow the copter fleet.  One of the things that was most impressive to me was how tight they have been able to make the waypoint tracking, even at speed.  For turns less than 60 deg, the copter doesn't even slow down. Look at version 2.6 waypoint tracking (top) vs. 2.7.3 (bottom):

Black Box: GPX -Viewer

Using the GPX-Viewer we are able to see the actual flight taken by the copter.  This was a transect we threw together while out flying today.  

Another feature of this program allows you see how things vary as one scrolls through the flight path.  This should help us identify why problems are occurring.

I could see the copter's position varied about +/-1 m.  Interestingly, the ground speed fluctuated predicatively with position in the path.  Before the corners and traversing to the next transect the copter slows then in the straight legs the copter picks up speed.  This is something we see in the number of redundant images being taken at the "turn arounds".  

Updated Image Processing Workflow

In order to keep current, here is a photo of our updated image processing workflow. Now, we convert from JPG and JPGW into a GeoTIFF using the OPEN_EV program. The OPEN_EV program has a GUI and is very intuitive, just requiring an export. Currently there is no way to batch convert the files as GeoTIFFS, but it doesn't take very long (just a few clicks per file).

Once turned into a GeoTIFF, we need to set the projection. For a still unknown reason, I can't figure out why Palentier is not producing a proper projection file for the JPG when it matches the jpg and gpx files. This may be a flaw in the code/design. Setting the projection is easy - in Arc, just define the projection based on what is in the original jpgw file (through the toolbox). Batch 'define projection' is currently not working for me, so again we need to do this individually, but at least it functions.

Until we have focused imagery, it'll be tough to make a proper determination as to which mosaicking procedure works best for us.

DeaconEye Flight Algorithm

Below is our standard flight procedure, which we use for all of our routine imaging flights.

Prior to takeoff, we connect the battery and do a full systems test. Wireless serial connection to the MK Tool is used to confirm settings on the copter and do a motor test. The motor test via the MK Tool is done not only to confirm the function of the motors but also to confirm functional 2-way communication between the copter and computer. The location and heading of the copter as projected in the MK Tool OSD is compared to its known position and heading to confirm for accuracy. After the systems test the waypoint flight plan us uploaded to the copter and then downloaded back to the computer. The download is carried out to ensure that all waypoints were captured by the NaviCtrl. At this point the pilot is cleared to start his takeoff procedures.

The pilot calibrates the gyroscopes and the camera stabilizer servos are confirmed to be functioning. After calibration, the pilot arms the motors. After motor arm, the copilot reads the battery level and confirms that the camera is triggering. Finally, the copilot confirms that Position Hold (PH) and Care Free (CF) are activated, and that Altitude Hold (AH) is deactivated. Once these checks are completed the pilot is cleared for takeoff.

The pilot increases the throttle in order to bring the copter into a hover at approximately 3-5 meters. When the copter is in a stable hover AH is activated, followed by Coming Home (CH) activation. When CH is activated the copter begins waypoint flight. During flight the pilot monitors the copter through direct visualization while the copilot monitors the copter through the MK Tool OSD. The copilot reads aloud the flight status (LiPO voltage, waypoint number, ground speed, and altitude) throughout the flight and monitors for any abnormalities in the values reported from the electronic speed controllers (ESCs) or any deviations from the flight plan.

When the copter reaches the end of its waypoint flight plan, the copilot initiates the CH command from the MK Tool OSD, causing the copter to return to the point where its motors were activated. The copter will fly to this position at the predetermined CH altitude. Once the copter reaches this point the pilot places the copter back into position hold and disables AH. With AH off, the pilot throttles the copter down for landing.

After landing, the flight time and final battery voltage is recorded.

Check back later for more detailed information on how we log the details of our flights.

LiPo Objective Figure of Merit

There's a lot of marketing hype in the "C" rating of LiPo's.  The C-rating is the highest recommended discharge current from a cell.  Manufactures have been over estimating their C ratings for awhile.  So people have created tools addressing this issue.  One that I like normalizes the C rating using the internal resistance (IR) and cell capacity.   This tool is called the LiPo Objective Performance Calculator: LIPo Tool One wants cells that have the highest capacity with the lowest internal resistance.  Our Quadropower 4s 6200 mAhr have an average IR of 1.5 milli-ohms (really good value).  So from the web site we see:

Figure of Merit - 1.21 (the higher the better)
Max Current = 152 A or ~20 amps per motor
True C rating = 25

As the cells are cycled their IR's will increase.  When the IR = 5 they will only be able to deliver 85 amps comfortably.  This is only ~10.5 amps per motor!!!  At this point the voltage of the cells will sag very low (V = IR, right?) and we will see the low voltage warning.  We have learned that knowing the health of your battery is fairly import parameter for controlled flight.  ~mww

Converting JPG to GeoTIFF

Here is a link that will help us with our JPG to TIF issues. It uses GDAL, which is a program that I'll put on the lab machine.

http://gis.stackexchange.com/questions/2312/how-can-i-transform-a-georeferenced-jpg-to-a-georeferenced-tif

Here is the text of the answer:

You could use gdaltranslate.
gdal_translate -of GTiff myFile.jpg myFile.tifThis will actually give you a geoTiff, not a tiff with a world file.

And for more information on GDAL:

gdaltranslate is one of the commandline utilty tools that are part of gdal [gdal.org/]. gdal is used by both OpenSource and ESRI (raster only) tools to provide support for reading/writing raster and vector data. For windows, the easiest way to install it is to either use the OSGEO4W installer or the FWTools installer. The OpenSource Desktop GIS QGIS has a plugin called GDALTools which allows you to do this using a gui tool.

I'll try it out and see what happens...

Here is the workflow for our image processing. This is as of 9/3/2012.