Support Website Contact Support Blog

How is the height determined from images and henceforth how is GSD computed??

I am confused on the Pix4D’s outputs for fisheye lens(GOPRO) images, No Geotaf. The pix4D result in the quality report for GSD and Area covered is Undefined. I would like to know how the GSD is computed. Is it dependent on the Geotag information of images

Anybody correct me if i am wrong but if you do not have geotags then the project will have no scale. It can build a model still because it is matching like points in each photograph and doing trig between the focal point of the camera and the point on the ground, so it will be relative. Until there are geotags it doesn’t know how big or small that project as a whole is in space. Thats why you would need GCP’s on the ground to provide that scale if the photos have no geotags. If the program does not truly know how big the project is it cannot give you an area or a GSD.

hope that helps

That seems like a fair explanation. But when I was playing around with the Geotagged images. I realized something, the GSD is computed from the Lattitude and longitude of the images but not from the height. Meanwhile I also came across a academic paper, wherein it was mentioned GSD = (height/focal length)*CCD Pixel size. I am attaching the paper for reference.

If it is determined by this formula, height plays a crucial role. Whereas, pix4D doesn’t abide to this. May I know if there is a relevant literature review on the same from pix4D?

Hello All,

The Ground Sampling Distance (GSD) represents how many centimeters in the reality are represented by one pixel in the model. For more inforamtion:

The GSD defines the accuracy and quality of the final results.

It depends on:

  • Sw = real sensor width [mm]
  • FR = real focal length [mm]
  • H = flight height [m]
  • Dw = distance covered on the ground by one image in the width direction (footprint width) [m]

For more infomation:

You can find a GSD calculator on this link:

More information:

The accuracy of your outputs depends on the accuracy of the inputs you give to Pix4Dmapper, i.e. the GSD, the accuracy of the image geolocation (coordinates of the initial images), the accuracy of the GCPs measured on the field and the visual quality of the images.

The accuracy can be distinguished into:

  1. Relative accuracy: It is defined as the difference between a distance measured on the reconstructed model and the real same distance. For example two points of the model can be 2 meters away from their real position on the earth but if their relative accuracy is high, then the distance measured between these points will be very accurate.
  2. Absolute accuracy: it is defined as the difference between the location of features on the reconstructed model and their true position on the Earth.

The relative accuracy of the outputs should be expected to be 1-2 pixels (GSD) horizontally (X,Y coordinates) and 2-3 pixels (GSD) vertically (Z coordinate).
GSD (Ground Sampling Distance) is defined as the linear ground distance that 1 pixel represents.
For instance, for a project of ground sampling distance (GSD) of 2 cm, the horizontal accuracy expected is about 2-4 cm and the vertical one 4-6 cm.

Using Ground control Points (GCPs) may improve the relative accuracy of your model (especially at areas with lower overlap or difficult image content).

The use of GCPs increases a lot the absolute accuracy of your project, placing your model to the correct position on earth.
The absolute accuracy depends on the accuracy of the measured GCPs, their number and their distribution.
The accuracy of your GCPs should be better than your GSD.
So, in the aforementioned example less that 2 cm.

If you do not have GCPs, then you need image geolocation to scale, rotate and orientate your model.
Then the absolute accuracy of the model depends on the accuracy of the GPS image geolocation, which is worse than the GCPs accuracy.

The Point Cloud, DSM and consequently the Orthomosaic accuracy are also affected by the quality of the initial images and their visual content.
Sharp edges, reflective surfaces, certain type or roads and rooftops may present locally lower accuracy.

More information here: