Lately i’ve been trying to wrap my head around Control points and how you can marge multiple components into 1 but for the life of me I cant figure out how they work. I watched the tutorial provided by RC on YT but I cant seem to get the same results as the YT video.
So far I get lets say 2 components, I add a few control points from matching images to align them but clicking update or align they still stay the same. Im most likely doing something fundamentally wrong but was hoping someone could give any more information.
Have you made at least 3 control points and confirmed them on at least 3 images in both components? (this is absolute bare minimum). If so what kind or error range are you getting on the control points? Also which videos have you watched so far, maybe I can point you towards some more.
Hi Wallace, ahh I wasnt aware I had to put the same amount and same place in both control points in my case. Regarding the error range im not entirely sure what that is so wouldnt be able to give a right reply. The video I watched was this.
Essentially when you make a control point you are telling the software “hey this thing in this image is exactly the same thing as this thing in this other image, and also this same thing in this image”
in the image above, the position where i confirmed “dsc05120.jpg” to be is 0.1 of a pixel different from where RC thought it was, this is actually called the image-measurement deviation not error range is i referred to it.
No 3 completely different features ideally in a triangle not a line (at least 3, you’ll need a lot more in a big scene). And each point needs to be aligned to at least 3 images in each component.
do this for all 3 control points. But there is no golden rule for how many control points and how many images in each one you would need to get them to merge.
Let me know if you have any more trouble with it. It can be a bit tricky at first, but once you have done it once it’s quite easy unless you have a really bad dataset.
Hi, I would like to have a little more info on this.
I have large structure where I used lot of CPs, I must put some more to finalise it. How does weight actually work. I have some CPs floating around structure, is this correct or this CP is not ok. When I change weight it moves. Can you explain logic behind this movement?
Hi Rango986, weight defines the influence of a point in calculations. RealityCapture uses natural features and control points together. In general, there are much more natural features than user-defined control points. If some of these points are matched incorrectly, then it is difficult to align images correctly. The weight is one way of saying “use this point and drop all natural points which do not agree”. On the other hand, increasing the point weight will make the alignment sensitive to errors in positioning of the point in the image. Even with the best efforts, it is difficult for a human to place a point with a precision of under 3-4 pixels, while natural points are placed with sub-pixel accuracy. However, it is recommended to leave the default weight value.
By changing this value you are setting the higher priority to that point. If the value is higher, the position is taken as more serious for the application. The changing of weight is recommended, where you are sure that the point is on that place and you want to use it in the alignment. If CP is floating around the structure, it is not good and it should be placed more preciously or not used at all.
If it is everything OK, then those CPs won’t be in the air. Is there some error for those CPs in 1Ds view? The result could be also bad alignment of some images. On how many images are those CPs placed?
At least 3, usually i go with more. And it is the same problem with more CPs, but I should mentioned that images were taken with Dewarping ON . I guess that could mess it up?
Basically wrong alignment for images and incorrect calibration parameters. Using those wrong information were the CPs computed with a wrong position in the space.