Forum

Author Topic: about photo alignment  (Read 4038 times)

NR1168

  • Newbie
  • *
  • Posts: 30
    • View Profile
about photo alignment
« on: April 09, 2016, 03:53:33 AM »
Hi, why my photo alignment is look like attached photo... H overlap 80%, w overlap 60% and it works for other field view. But for this field it didn't work. What is the problem? and What should I do?
« Last Edit: April 09, 2016, 04:08:37 AM by NR1168 »

gatsri

  • Jr. Member
  • **
  • Posts: 77
    • View Profile
Re: about photo alignment
« Reply #1 on: April 11, 2016, 08:16:40 AM »
What is the vegetation? in some cases there is more overlapping recommended. try to play with the align settings. (lower accurancy, lower Key points limit) in some cases it helps to get these images aligned...

NR1168

  • Newbie
  • *
  • Posts: 30
    • View Profile
Re: about photo alignment
« Reply #2 on: April 15, 2016, 09:34:41 AM »
Thanks for your reply. Now there is very less vegetation.... just finished tillage... but not working... H 80% and W 60% is the overlap ratio.

NR1168

  • Newbie
  • *
  • Posts: 30
    • View Profile
Re: about photo alignment
« Reply #3 on: April 15, 2016, 09:47:16 AM »
I lowed down the accuracy and Key points  but still not working

HMArnold

  • Jr. Member
  • **
  • Posts: 52
    • View Profile
Re: about photo alignment
« Reply #4 on: April 17, 2016, 05:11:17 PM »
From what I have learned so far, bare fields are more difficult for camera alignment than fields with crops. I have always assumed it's because there are fewer distinct points for Photoscan to tie together.

I'm attaching the results of an agricultural survey mission at 100M over a field that had just been planted the day before - essentially a bare field.

On that same day I flew a total of 900 acres with a Phantom 3 Pro, and all fields had complete image alignment.

With a P3P using the Android Ultimate Flight App, you can export a detailed mission flight log in CSV form that includes the YAW, or platform heading, then create a Photoscan import file that includes that value for every image.

If you include the heading it greatly increases the chances of complete camera alignment.

Also, the altitude values from the EXIF info of a DJI P3P are from the GPS altitude, which can be off by 50 or more meters from reality, but since they appear to me to be stable across a single mission, they are accurate relative to each other. That means that the relative altitude of each picture is good if all you want is an orthophoto, but the altitude of the entire field if you export a geo-referenced tiff can be off by 50 meters. This is easily fixed by one or two synthetic GCPs with altitudes derived from Google Earth.

With accurate relative altitude and heading values, you can lower the "Accuracy" value for each camera definition. As I understand, lowering the "Accuracy" tells Photoscan to rely more on what they call the "orientation" values, which are altitude, heading, and gimbal angle.

With the default "Accuracy" value of 10 meters, I think Photoscan wanders about turning and trying different altitudes with images for some allowed amount of effort, then gives up and goes to the next image. That's why cameras that align and camera that don't seem to group together - Once Photoscan has a good match the next image fits right in. Once it's lost, twisting and turning the next image doesn't very often get it back on track, so it remains lost.

Since most agricultural survey grids are composed of parallel lines, there are really only 2 YAW values you need to enter, which can be done by hand. For images taken during the turns, if you can't include the platform heading at that instant, I suggest you take them out, but from your alignment results it doesn't look like you're having trouble with the grid ends.

At the same time, set the PITCH (gimbal angle) value to "0" because the UAV keeps the camera pointed straight down, and the ROLL (UAV lateral angle) value to "0" because the camera is gyro stabilized.

After entering all these values, I set my "Accuracy" to 1 meter for all cameras.

You will also notice in the attached page from the report (attachment limit exceeds entire report file size) that my overlap is significantly higher than your 60% horizontal.

In my experience the vertical (along the flight path) is not near as important as the horizontal (between grid lines) distance.

The bad news is that it's the horizontal distance that defines how many acres you can fly using a single battery... The closer the grid lines, the better camera alignment results. The further apart the grid lines are the more acreage you can cover but it doesn't do you any good because you can't produce an orthophoto.

My suggestion is that you fly that same field again with a closer gird pattern, create an import file that includes the YAW, PITCH and ROLL values for each image, and lower the accuracy for each image to "0".

If that mission creates acceptable camera alignment, try another mission done the same way with slightly wider grid distances (horizontal overlap) until it fails again, then remember that's the grid width you need for bare fields.

The good news is that once the crops start to grow, you will find that you can get alignment with significantly wider grids.

Hank
Texas

NR1168

  • Newbie
  • *
  • Posts: 30
    • View Profile
Re: about photo alignment
« Reply #5 on: April 18, 2016, 03:45:57 AM »
Thank you very much for your suggestions....
I will try it..