|
libfovis
|
Provides depth estimates for input image pixels. More...
Public Member Functions | |
| virtual double | getBaseline () const =0 |
| virtual void | getXyz (OdometryFrame *frame)=0 |
| virtual bool | haveXyz (int u, int v)=0 |
| virtual void | refineXyz (FeatureMatch *matches, int num_matches, OdometryFrame *frame)=0 |
Provides depth estimates for input image pixels.
A reliable source of depth estimates at as many pixels in an image as possible is crucial to producing useful visual odometry estimates. The DepthSource class abstracts out the process of determining depth at each pixel.
DepthSource is used in a lazy fashion. First, keypoints in the input image are identified. Next, keypoints without depth information are filtered out (pixels where haveXyz() returns false) and discarded. Later, the actual depth of the remaining keypoints is retrieved using getXyz(). Finally, after features are temporally matched across frames and their keypoint positions refined, DepthSource is used again to refine the depth estimates of the refined keypoint image positions.
|
pure virtual |
This should return true if it's not certain there's no depth at (u,v). It should be an inexpensive check that is used to avoid pointless (hah!) creation of keypoints. False positives are fine as ling as getXyz gets rid of them.
Implemented in PrimeSenseDepth, StereoDisparity, StereoDepth, and DepthImage.
|
pure virtual |
Populate keypoints in frame with XYZ data.
Implemented in PrimeSenseDepth, StereoDisparity, StereoDepth, and DepthImage.
|
pure virtual |
Refine XYZ data of target keypoints in matches (usually after subpixel refinement of the matches across time).
Implemented in PrimeSenseDepth, StereoDisparity, StereoDepth, and DepthImage.
|
pure virtual |
Return baseline of depth source, if applicable. If not applicable (e.g. for OpenNI devices) return 0.
Implemented in PrimeSenseDepth, StereoDisparity, StereoDepth, and DepthImage.
1.8.9.1
