Topics

1

General / Zero Resolution issue

« on: April 27, 2022, 11:30:32 PM »

Greetings.

I am running Metashape version 1.8.3. build 14331 (64 bit). I am processing imagery from a Micasense dual camera system. It is a fairly large project with 2869 cameras. I've processed jobs of this size in the past with no problems.

I'm running into an issue that I've never seen. When I get to the point of generating a dense point cloud, it chugs for a bit but then gives me a "Zero Resolution" error. I found the note regarding this issue that suggests that updating the driver for my GPU may solve the issue. I've done this but I still get the same error.

The bounding box is quite a bit larger than the extent of the sparse point cloud (another thing that is mentioned as a possible fix).  In the photo panel, clicking on the depth map indicates that none of the images have a depth map but I'm not clear on what this means.


Specs on my computer:

Intel(R) Xeon(R) CPU E5-1620 v3 @ 3.50 GHz
RAM 32 GB
GPU: Nvidia GeForce GTX 1080 Driver version 512.59

I'd be grateful for any suggestions.

David Wallin

2

General / Sequoia camera; reflectance calibration without reflectance panel images

« on: October 29, 2021, 02:26:58 AM »

Greetings. I have some old images acquired using a Parrot Sequoia 4-band camera. Unfortunately, it appears that either I neglected to get images of the reflectance panels or maybe I have just lost those images. I did fly with the sun sensor in place. I tried doing the "calibrate reflectance" procedure just using the sun sensor but I got a "calibration failed" message. Can anyone provide a suggestion on how to proceed? It is not clear to me if the sun sensor data is attached to each image or where it is stored. In looking at the image properties and "details" tab for each image, I don't see anything that looks like calibration data.

Thanks for any suggestions!
David Wallin
david.wallin@wwu.edu

3

General / Red_edge_dual_dense_pt_cloud_problem

« on: June 17, 2020, 01:04:26 AM »

Greetings.

I am processing imagery from the Micasense Red Edge Dual Camera system. This generates a 10-band image. The data set includes 2170X10 tiffs covering a study area that is about 250 m by 2500 m. I have successfully run the reflectance calibration procedure and all of the images aligned. I have 28 GCPs. The sparse point cloud looks fine (jpg attached). When I run the dense point cloud procedure, the result has large holes in it (see attached JPG).

Any suggestions on how I can get a dense point cloud that covers the entire study area?

Regards,
David

4

General / align thermal images from XT2

« on: September 04, 2019, 07:52:59 AM »

Some time ago, I posted a request for guidance on aligning thermal images from an XT2 camera (512X640 resolution 13 mm lens). I've received no suggestions so hoping that by reposting I might get some suggestions. I have followed the same workflow that I've used for aligning standard RGB images but when doing so, only 1/4 to 1/2 of the images align.

I would be grateful for any suggestions.

Regards,
David Wallin

5

General / thermal mosaics using XT2 camera

« on: May 31, 2019, 12:41:35 AM »

Greetings. I just purchased a DJI Zenmuse XT2 thermal camera; 640X520 resolution with a 13 mm lens. I'm flying it on a DJI M210 quadcopter.

My objective is to create thermal mosaic images (and a co-registered RGB mosaic and surface model) of rivers to detect thermal refugia created by intermittent seeps of cold water from small tributaries and springs.

Can anyone point me towards a workflow for achieving this? Recommendations for mission planning software as well as image processing recommendations would be greatly appreciated.

Regards,
David

6

General / Camera calibration; importing data from Lens software

« on: December 11, 2018, 09:28:59 PM »

Greetings,
I am using a DJI x4s camera on an M210. I ran the Lens software to acquire camera calibration data and I've then imported these data into Photoscan. I'm confused since the parameter values displayed in the Lens software do not match the same parameter values after importing into Photoscan. See the attached jpgs.

Can anyone explain this?

Regards,
David

7

General / placing markers in Parrot Sequoia imagery; green images too dark to locate GCPs

« on: December 05, 2018, 01:57:14 AM »

Greetings. I've just aligned a set of multispectral images acquired using a parrot sequoia camera. In the photo panel in Photoscan, it is just displaying the green band for each image (not the red, red edge or NIR bands). The images for the green band are all very dark and it is extremely difficult, if not impossible, to see the GCP markers that I placed throughout the study area prior to the flight. Is there a way to display the other bands or can I somehow brighten the green images to facilitate the placement of markers in the images?

Thanks,
David

8

General / parrot sequoia calibration

« on: November 30, 2018, 02:48:21 AM »

Greetings. I am processing Parrot Sequoia multispectral imagery using the "Radiometric calibration using reflectance panels in PhotoScan Professional 1.4. I have images for the reflectance panel and I've located the panels as indicated in step 3 of this tutorial. I am using this calibration panel for the first time. This is the calibration panel that came with the camera. I don't have a "calibration certificate" that is mentioned in the tutorial. Not sure if one came with it or not. In the "Calibrate Reflectance" dialog box, it is showing a reflectance of 1 for each of the 4 bands. This seems unlikely.


Am I doing something wrong here? Where do I get the correct reflectance values?


Isn't another option to use data from the sun sensor and if so, how do I do this? I've brought in the tiff files that represent green, red, red edge and NIR, but not the jpg files that contain the 16MP RGB images. The folder also contains two .dat files (sequoia_diag.dat and sequoia_param.dat). Do these contain sun sensor info?


Any help would be much appreciated.


David

9

General / image alignment of forested areas

« on: November 27, 2018, 01:52:31 AM »

I realize that the alignment of images from forested areas is challenging. I'm wondering if anyone has any tips for getting more images to align. Specifically, in the "Align Photos" dialog, I'm wondering if using a lower accuracy or changing the Key point or tie point limits would help (and what settings for these options might help).

Thanks!
David
 

10

General / Image overlap layer from Processing Report

« on: March 12, 2018, 06:47:49 PM »

Is there a way to get the Image overlap written as a tiff file? I'd like to be able to use this to clip both my mosaic and DEM to eliminate areas with a low number of images. I'm attaching a copy of this layer that appears in the processing report. Photoscan obviously generates this so I'd think that I should be able to get it but not sure how.

Thanks!
David

11

General / Fusion of precision geotagged images with other images

« on: January 08, 2018, 11:44:24 PM »

Greetings.
I'm considering the purchase of an RTK GPS system like Vmap:
http://www.microaerialprojects.com/

This system requires the use of a camera with a hotshoe (ex. Sony A6000) to link individual images with the precision GPS coordinates of the Vmap. Nice but, I use several different cameras that lack a hotshoe (near-IR modified Canon S100s and Parrot Sequoia).

So, I'm wondering if there might be a way in Photoscan to fuse imagery from the Vmap-linked RGB camera with, say, a Canon S100 camera (or another camera with either no geotags or geotags with low accuracy) to create a multiband image that uses the high-accuracy geotags from the V-map linked camera. The goal here is to avoid the time consuming process of setting up ground control points and then finding these geotags in the imagery.

David
 

12

General / Vertical error in DEM and Ortho

« on: January 08, 2018, 10:36:59 PM »

Greetings,
I've created a ortho mosaic and DEM using 2300 images and I'm seeing large (+-30 m) vertical errors in the resulting DEM and associated visual anomalies in the ortho.
 
This is for a mapping project for eelgrass in a nearshore marine environment. I'm flying a ~300 by 2500 m transect long-term monitoring site at an estuarine research site. We flew it in a series of 500 m sections over a two-day period. We had to time the flights to correspond to extreme low tide events and each day of flying took place over a window of about 4 hours. The logistics of pulling this off were challenging. We did flight with both near IR modified Canon S100 cameras and a Parrot Sequoia.

Along the transect there are PVC pipes pushed into the mud to mark plot locations. We put large marker panels that would be visible in the imagery over some of these PVC pipes and obtained GPS coordinates for these locations. The plan was to use a high resolution RTK GPS unit to obtain precise XYZ coordinates for each GCP. There were problems with the unit and the resulting coordinates were of questionable accuracy but I figured the XY coordinates would be good enough to get started. Along the entire 2500 m transect, the elevation change is minimal (about 1-2 m) and I really don't need a DEM anyway. So I just gave each point an elevation of zero. I had about 40 GCPs. I retained about 30 of these (with the lowest error) and used the others (with higher errors) as check points. For the S100 images, total error for the control points was about 3 m and about 1.5 m vertical error. But more like 20 m total error and 14 m vertical for the check points.

After separately processing both the S100 images and the Sequoia imagery, I'm seeing what I regard as very significant vertical errors (+-30 m)  in the DEM and associated visual anomalies in the ortho.  Although I was expecting a fair amount of XY error, I figured that, since all GCPs had an elevation of zero the entire study area would end up dead flat.

Other than getting better coordinates for the GCPs, is there any way to fix this?

Processing report is too large to attach so I've attached a subset of this report

David

13

General / Creation of a 5-band image from two IR modified Canon S100 cameras

« on: January 03, 2018, 08:55:43 PM »

Greetings. I have two modified Canon S100 cameras. One covers the blue, green and red edge, the other covers the green, red and near-IR. I fly both of these at the same time on a quadcopter with each triggering independently at a 2 second interval. This means that images are not acquired at the same instant. Transect spacing and camera trigger interval set to provide lots of overlap on all images. I use Mission Planner to create the flight plan.

I'd like to fuse these images to create a 5-band multispectral ortho covering the blue, green, red, red edge and near IR.

Can anyone provide me with a workflow for doing this?

Thanks!
David

14

General / Dealing with incorrect marker placement

« on: November 20, 2017, 03:53:09 AM »

Greetings. I am using the Orthophoto/DEM with GCP tutorial. I've aligned my images and I am placing markers. After finding a marker in a single image, I filter photos by marker. I then go through the images one at a time to refine the marker location. I have run into two different situations where the marker location is incorrect.

1. I some cases the view of the marker is blocked by a tall tree but the marker position is more or less in the correct location in the image if the tree wasn't there or if you could see through the tree. So in this case, I right-click on the marker and select "Remove Marker."

2. The other case which is less common is when the marker is place on an image that definitely DOES NOT cover the position where the marker is located. On the Model tab, this image is clearly in a completely different part of the study area. In this case, I also right-click on the marker location and select "Remove Marker."

Am I doing the right thing in both cases?

Thanks for your help.

David

 

 

15

General / Aligning Sequoia RGB images

« on: August 24, 2017, 09:32:48 PM »

Greetings. I have a set of 3567 RGB images acquired using a Sequoia camera. The images were acquired along a ~200 m by 2500 m transect on 10 flights over a two day period using a 3DR Solo. We were flying over exposed eelgrass beds at low tide. Nearly flat terrain. Nearly all of the images were geotagged by the Sequoia camera (a handful of images lack geotags).

I am using Agisoft Photoscan V1.3.2 build 4205

I am having difficulty aligning the images. I've done it twice, once using only the "Reference Preselection" checkbox (Under "General" on the "Align Photos" dialog) and a second time with both the "Generic preselection" and the "Reference preselection" boxes checked. I am using the default values for both Key points and Tie points.

In both cases, the results are poor: 2114 of 3567 images aligned using just Reference preselection and 1979 or 3567 images aligned using both Generic and Reference preselection. Images of both alignment results are attached. Both show quite dramatic (and clearly incorrect warping of the image positions. The quadcopter was holding altitude reasonably well throughout all flights (probably plus or minus 2 or 3 m; flying at 60 m elevation). I was using sidelap and frontlap of about 80%.

Any ideas how I can fix this to get my images to align?

Thanks!

David