Topics

1

Python and Java API / Batch script produces high error OR distorted ortho/DEM

« on: August 15, 2018, 05:08:03 PM »

Hello,

We have been using a batch processing script that was updated from ver 1.1 to 1.4. Our original script provided a beautiful ortho/DEM but the target error was quite high (>1 cm on a small scale object), which produced an overall 'doming' to the features.

By making the camera calibration automatic, the error was greatly reduced (2-3 mm) but the ortho/DEM are distorted or warped on the edges.

We have a dense target network and the only difference between the scripts is that the use of precalibrated camera parameters or automatic calibration (lines 61-67 in attached).

We are hoping that we don't need to choose between error OR distortion. Any help would be appreciated. Also, we have tried splitting the two cameras used into separate auto calibration but this did not improve error or distortion.

Thank you

2

Python and Java API / Doming with target detection

« on: June 04, 2018, 07:40:41 PM »

We are using an updated automatic batch processing script to process large amounts of data. We have success with target detection but the errors on the targets are huge. We are also experiencing the well documented 'doming' effect.

Moving from version 1.1 to 1.4, I want to make sure we are 'optimizing' the photo alignment and cameras properly. After photo alignment and target detection, we are using 'optimizeCameras'


Match photos (perform image matching for the chunk) and align camera (perform photo alignment for chunk)
    chunk.matchPhotos(accuracy=PhotoScan.HighAccuracy, generic_preselection=True)
    chunk.alignCameras()

    doc.save()

### Detect markers
    chunk.detectMarkers(type=PhotoScan.CircularTarget12bit, tolerance=80) 
    chunk.loadReference(coord_path, format=PhotoScan.ReferenceFormatCSV, skip_rows=1, columns="nxyz",
                        delimiter=",")  # you can alter the columns order and delimiter according to the input csv format
    chunk.updateTransform()
    chunk.marker_location_accuracy = PhotoScan.Vector(
        [0.001, 0.001, 0.001])  # you can input different accuracy along every axis

   
    print('Fini les cilbes debut seuil')
    doc.save()
    print('debut optimisation')
### Perform optimization of point cloud/camera parameters
    chunk.optimizeCameras()

Help is greatly appreciated!

3

Python and Java API / Updating Batch Processing Script from Agisoft 1.0 to 1.4

« on: March 09, 2018, 11:42:24 PM »

Hello,

I am trying to update a python script that was used for batch processing large numbers of photos in Agisoft 1.0 to Agisoft 1.4.

I am working with a dataset (photos, target locations, and camera calibration file) that was run successfully through the old version - so the only difference is the version of Agisoft!

I am not a coder, and I did not write this code so any help would be greatly appreciated.

The original code:
doc = PhotoScan.app.document
        doc.clear()
        chunk = PhotoScan.Chunk()
        chunk.label = "New_Chunk"
        doc.chunks.add(chunk)

TypeError: cannot create 'Chunk' instances



When I force it past that point the next problem comes from the original script:

calib_in = PhotoScan.Calibration()
        calib_in.load(CalibFile)
        try:
            sensor = chunk.sensors[0]

IndexError: list index out of range

Thank you in advance.