Interpreting Confidence Values in Dense Point Cloud

[m.lec]

Hello,

I just switched from the free version to a trial of the full 3DF Zephyr and am exploring the different features that come with it, the Confidence Tool being one. I read the tutorial on using the Confidence Tool, but am curious about how to interpret the information so that I can better understand how to move forward in processing and filtering.

For example, is there a threshold value above which you should remove points with low confidence as bad data?

Is there an ideal confidence range? Is a lower number better than a higher number, or vice versa?

What does a value of 0 mean?




My charts tend to skew to the left, I included an example below.



Thank you!
Miranda

[cam3d]

Hi m.lec -

As I understand it, confidence in a dense point cloud is determined by the number of images across which a point is correlated.

Higher confidence values indicate more reliable points. Points with low confidence, closer to 0, are less reliable and may be a result of noise in the data.

• Higher Confidence: More reliable points (matched in more images).
• Lower Confidence: Less reliable points (matched in fewer images).
• Value of 0: Highly unreliable or undetected points.

Filtering out points below a certain confidence threshold can improve the quality of your point cloud.

Hope this helps!

[Andrea Alessi]

Adding one comment on top of what Cam said, is that the confidence values can't be compared within two separate datasets.

The confidence score is an arbitrary number as it takes into account multiple factors. They are used within the same dataset to sort images / remove points below certain thresholds, but it's important to remember that a score of X on a dataset, doesn't mean it's necessary worse, or better, than a score Y in a different project.