Re: RGBD SLAM and then RGB Localization

Posted by Kaifei on
URL: http://official-rtab-map-forum.206.s1.nabble.com/RGBD-SLAM-and-then-RGB-Localization-tp489p523.html

Thanks! I changed my code to use the PnP codes in Memory.cpp.

However, even though I can find a similar image in the database, I often cannot find enough inliers.

For example, for the two images I attached, there are only 12 inliers, among which 0 is PnP inlier. (for some other images I can have 5-10 PnP inliers so I can calculate the transform)

These two images look very similar to me, and they seem to have many features.
So I'm wondering whether it's normal that there are not many inliers?

Based on my understanding, a word has to be unique in both images to be considered as an inlier.
It seems that I can increase the number of unique words by decreasing _nndrRatio in VWDictionary.cpp.
But I'm not sure whether that's a good idea?

Also, I'm using Kinect v1 to collect RGBD data.
Will using Xtion PRO LIVE increase the number of inliers?

Thank you so much!