Support Website Contact Support Blog

Importing Intrinsic

Hello,

I took some images from a camera for which I dont know what the model is. So, I have calibrated it myself using a chess board and obtained the intrinsic matrix.

1) When I attempt to process these images in pix4d, it asks me for camera parameters which I don’t know. Is there a way of directly importing the intrinsic matrix

  1. When I try to edit the camera model, I cannot change sensor width and height together. My images have an aspect ration of 4:3 but the current settings in sensor width and height seem to be 16:9 which cannot be changed

Any help on above two issues is highly appreciated

 

Thanks

Mahanth

Hello Mahanth,

External software used to calibrate cameras is not always compatible with Pix4Dmapper.
The way of defining the internal parameters may differ this is why we recommend to calibrate your camera directly using Pix4Dmapper.
Moreover our software is able to detect more tie points between images which makes the calibration more reliable.

The procedure to calibrate a perspective camera is as follows:

The camera calibration using Pix4Dmapper relies on the processing of a good dataset (check whether your current inputs could be used).

Ideally it should be a 100 images aerial nadir data set with sufficient overlap of at least 75% frontal overlap and 60% side overlap.
However the optimal overlap depends on the type of survey and terrain.
For more information, please check: https://support.pix4d.com/hc/en-us/articles/202557459
In case of terrestrial or free flight for instance, we suggest that you increase the overlap to 85% and 70% respectively. The main issue is to avoid capturing sky background since tie points are less easily matched.

Once you imported the images into Pix4Dmapper, open the camera model editor and change the initial values of the internal parameters.
Before inserting any values, click the Clear button. Then you just have to manually insert the Pixel Size in μm and the Focal Length in mm or pixel. The image/sensor dimension is directly retrieved from the content of your inputs. In addition, the distortion parameters will be fixed during processing so make sure to initialize them to 0.

How to create a new camera model: https://support.pix4d.com/hc/en-us/articles/202560169

You are now able to start running step 1. Once it has completed, you should check the reprojection of 3D points computed by Pix4Dmapper.

Click some points and evaluate their location on the images (right sidebar). If you find some inconsistencies, we suggest that you add few manual tie points that should be marked on two images at least each. Don’t forget to click Apply every time you add a new point and click Reoptimize (Process menu) to improve the calibration.
Once it has completed, check again the reprojection of several 3D points.
Add manual tie points and reoptimize the project as long as you are not satisfied with your results.

Procedure to add and mark manual tie points: https://support.pix4d.com/hc/en-us/articles/202560349

Then you should replace the camera initial values with the optimized values: https://support.pix4d.com/hc/en-us/articles/202560059
Once you did, run step 1 a second time (not only Reoptimize).

Again check the reprojection of some 3D points, adjust your model if needed using the Reoptimize option, load your optimized camera values, and run step 1 a third time, etc. Repeat the steps so that the optimized internal parameters stabilize.
The last time, just run the project without the manual tie points (MTPs) you marked to check whether the project can calibrate well only with the optimized camera parameters, i.e. without any user intervention (MTPs).

Regards