local image matching returning duplicates and distant "matches"

[danbjoseph]

the light blue dots are the sampled points, the white camera icons are the locations of the images (the screenshot is of JOSM, a program that by chance makes it drag-and-drop to display the location of geotagged images on a map). the OSM roads extract i used, and therefore also the sampled points, extended beyond the area for which I captured imagery. the orange circle is around the location of image with filename GS010002_0_002510.jpg

(for this screenshot i was using QGIS, a FOSS desktop GIS software, which is why it looks different) the yellow dots all have data/images/ParkView/GS010002_0_002510.jpg as the assigned image. we shouldn't have duplicates like this and they are all some distance from the actual location of that image.

part of the solution may be to configure a sampling distance as part of #38 that is used in the create_points process (instead of passing the --mini-dist argument to set a distance in meters between interpolated points. then we can use that value in the imagery fetch step. we could divide it by 2 and use that distance as the buffer distance to search for matches. that is, if using the suggested distance of 20m between points, we would only consider images within 10 meters of a given point for a match. if we do this, i think there is still something off with the matching because of the duplicates we're seeing and the distance between the matches and the actual image location.

[danbjoseph]

i ran this after @dragonejt made some changes and it now returns as expected (only the green dots that match where imagery is present) and with no duplicate images. we should still take a close look at the image matching/selection process as part of #81