I thought I’d share a figure from last week’s science team call that the science team discussed. Michael was looking at combining clustered features with Shapely, a Python package for manipulation and analysis of planar geometric objects. Partly this is to investigate whether this could be used to deal with differing clusters in the overlap regions between neighboring subject images and also test out if we can use the software package to easily calculate the total area covered by the seasonal fans and blotches. Shapely does a good job of merging the blotches together as you can see from the figure below. This definitely looks like a way forward for calculating the total surface area per time of year covered in dark fan material.
Thanks to your help, we’ve finished search area two for Planet Four: Ridges. We’re working on analyzing the results and hopefully starting work on a paper based on those results. Laura has come up with a new region and slightly different type of polygonal ridge to search for. We’re working on getting that dataset processed and uploaded to the site. We hope to have this completed by the end of September with updated tutorials. We’ll keep you posted. In the meantime, Planet Four and Planet Four: Terrains could use some help if you can spare the time to classify an image or two.
We’re now working on dealing with the last major component of the Planet Four data processing pipeline, the overlap regions of neighboring subject images. We divide each HiRISE images into many smaller 840×648 pixel subimages or subjects that we show on Planet Four. To make sure we capture fans and blotches that are the edges of our subject images, we have a 100 pixel overlap between the neighboring left and bottom subject images. This means that we have duplicate markings that cover the same source which we need to identify as being the same source to allow for counting the number of seasonal sources and to also accurately measure the shape of very large fans or blotches.
We spent part of the last science call looking at some examples of overlap regions and the outputting fan and blotch shapes after clustering to decide what to do.
If you focus on the center sources in the two plots above, you see there are lots of markings identifying the same shape from the different subject images that contained varying parts of the central blotch or fan. Based on what we see, we think we if in the overlap region we only keep the largest source and anything that extends beyond that we will accurate identify the fan or blotch being marked. We’re going to test that this week and review the output from the catalog for a small portion of the overlap regions to confirm.
Once we sort what to do in the overlap regions, the focus should be writing all of the steps in the processing pipeline into the paper draft.