Maximum Sigma Naught and Achievable Block Adjustment Accuracy in Pixels

[Jake]

Hi All,

I am interested in understanding how ArcGIS Reality Studio performs in evaluating the quality of tie points or adjustment results during Aerial Triangulation. In software such as Trimble Inpho, we commonly rely on metrics like "High precision tie point correlation = 0.1 pixel" to gauge accuracy

.

Could you please provide information about a similar metric or capability in ArcGIS Reality Studio, specifically in relation to sigma naught? Additionally, I would like to know the maximum value of sigma naught or the block adjustment accuracy achievable in pixels.

Thanks,

Jake

[cam11315 · Accepted solution]

Hi Jake,

Thanks for reaching out. I can point out a few ways in which Reality Studio allows you to assess the quality of the Alignment (Bundle Adjustment):

• Sigma Naught is meant to be used as a broad quality indicator and it's recommended you consider other aspects of the results as well. The expected value varies depending on the camera system configuration (Nadir vs Oblique). For a Nadir dataset captured with a metric camera system, a value below 0.5 is a good indication.
• Reprojection Errors will give you an indication of the relative accuracy. They should ideally stay below 1 pixel.
• For improving absolute accuracy, it's important to use well distributed Ground Control Points. You can assess the absolute accuracy by using some Control Points as Check Points and inspecting their 3D residuals after the alignment is processed.
• It is also important that the images are well connected through tie points. You can check this by looking at the number of automatic tie points per image. This information is available in the Overview tab of the QA tool.

I hope you find this useful. Let me know if you have further questions.

Best,
Camilo