The objective of the project is to develop an algorithm to normalize the footprint of building polygons by eliminating undesirable artifacts in their geometry.
After the building mask is extracted from the satellite image via semantic segmentation, the building shape is often irregular. The objective is to regularise the building shapes via the algorithm developed. The algorithm should take care of only the regularisation/polygonisation part. The segmentation part is already completed. The algorithm has to be developed in Python and should not use any paid API calls.
The solution may use the polyline compression algorithm to correct distortions in building footprint polygons created through semantic segmentation that may produce undesirable artifacts. The developed solution should not use standard approaches, e.g. the Douglas-Peucker algorithm, which are greedy in nature.
The solution may use some priori building properties, which can be manually set. Some properties for example can be:
1. The building edge must be of at least some length, both relative and absolute, e.g. 3 meters
2. Consecutive edge angles are likely to be 90 degrees
3. Consecutive angles cannot be very sharp, smaller by some auto-tuned threshold, e.g. 30 degrees
4. Building angles likely have very few dominant angles, meaning all building edges are forming an angle of (dominant angle)
To summarise, the solution can follow any logic to output regularised polygons with min no of vertices. The regularisation process should take less than a sec for each polygon., in total for a large area with 10,000 polygons, the process should not take more than few mins.
At the time of bidding, kindly explain your algorithm in brief (just the building blocks so that I get a rough idea where we are headed). For solution acceptance, a processed mask (building footprint) will be shared as a numpy file and you will have to share the corresponding regularised mask along with time statistics.
About the recuiterMember since Aug 14, 2017 Terry M.
from Zhejiang, China