Can coded targets be enough to align images?

Answered

Comments

4 comments

  • Avatar
    Tom Foster

    Good question

    0
    Comment actions Permalink
  • Avatar
    Erik Kubiňan CR

    Greetings,

    Yes they should create a component and the more photos you add the clearer will the component be.

    0
    Comment actions Permalink
  • Avatar
    Tom Foster

    Great. Could you clarify:

    By 'more photos you add', does that mean each marker appearing in a larger number of photos? i.e. more overlap.

    By 'clearer will the component be', what does that mean? either a valid component is formed, or not - it's binary.

    This is valuable, because placing markers may be the best way to align photos of building interiors with featureless wall areas. Some idea of optimum (or sufficient) density of placed markers, and photos overlap, would help, alongside experimentation. This abstracted question, 'empty scene with nothing but coded targets', should help to clarify.

    0
    Comment actions Permalink
  • Avatar
    Erik Kubiňan CR

    Hello Mr. Foster 

    Yes of course the more pictures under different angles you add the better is the overlap as well as the density and quality of the result point cloud. You are providing information about the detail via images to RC which tries to find features in your images and matches them with other common images. The more information you give the more you get in terms of quality and accuracy. This doesn't obviously mean the scaling with this method is infinite but in some cases there may be too few images to get the correct results. Finding features also depends on the object shot so the more detail and the bigger the size of your marker or object is, the better. One more thing which is important is the framing of your subject. Keep in mind that you only use as many effective pixels of your image as how well you frame the subject while shooting.

    0
    Comment actions Permalink

Please sign in to leave a comment.