How to access the nearest neighbours found by a cv::flann knnsearch?

Question

Or is it even possible with flann ? Im not the most experienced coder, I also might just be overlooking something really basic (C++,OpenCV 2.4.3.)

The problem :

I have two pointclouds and want to calculate a displacement map. I am trying to use the flann .lib to get the nearest neighbour to a point in the first cloud from the points of the second cloud, and use them and the distance to calculate the displacement vector(s).

What I got so far is this:

int nn = 1;
cv::Mat MyIndex(data1.size(),3,CV_64FC1);
cv::Mat MyQuery(data2.size(),3,CV_64FC1);
cv::Mat indices(data2.size(),1,CV_32SC1);
cv::Mat distances(data2.size(),3,CV_64FC1);

cv::flann::Index_<double> NN_Index(MyIndex, cvflann::KDTreeIndexParams(4));
NN_Index.knnsearch(MyQuery,indices,distances,nn,cvflann::SearchParams(32));

It works as far as I can tell, I got the distances, I got the query points, I got the indices. But how do I get the actual points that got matched to my query points, from the indices ?

I looked through the flann.hpp but couldn't really find any hints. I messed arround a bit with MyIndex, NN_Index and the indices, but didn't get any useful results.

Answer 1

Try

for (int queryIdx = 0; queryIdx < MyQuery.rows; ++queryIdx) {
  int dbIdx = indices.at<int>(queryIdx, 0);
  std::cout<<"Query Idx:"<<queryIdx<<" matched to "<<"Database Idx:"<<dbIdx<<std::endl;
}