Skatespots are very likely to vanish sometime. May it be because they get closed using constructional elements like skatestoppers or they get torn down completely because they don’t fit in the plans of city development. Or they just get out-dated and need a re-construction.
I have long searched for methods to capture the complete skatespot using photography. I only got one special view. This works great for documenting the spots but there was no possibility to collect all the different transitions, gradients and edges, the roughness of the floor, the cracks and ripples - all that what transforms architecture into a skatespot. And this is what creates the feeling a skateboarder has when exploring the site.
I started to use SFM/MVS to create 3D point clouds from multiple photographs capturing different views of an object. Using these 3D point clouds it is possible to capture the skatespot in its entirety. I have experimented with this method before, but I didn’t find a suitable workflow based on free software that resulted in a high resolution point cloud. The software was either not free or too hard to handle and the computational needs have been too high.
The Colmap software and the availability of GPU processing power at reasonable price made it possible for me to use SFM/MVS for the skatespot reconstructions. Furthormore I’m using illustrative rendering of the 3D models to highlight structural elements and reduce the influence of superficial information like paintings or strong colors. These usually distract the viewer from the structures that I’m interested in.
I started research evaluating the methods and available software structure from motion / multi-view stereo reconstruction using free software only.
After a successful reconstruction of an example location using a workflow based on free software, I started to focus on a routine processing of the 3D reconstructions. Using the SFM/MVS approach, I changed from skatespot documentation to skatespot conservation using photography.
I got a dedicated computer for the computation of the 3D point clouds. It has two graphics card providing enough GPU resources to speed up my SFM/MVS workflow and to compute 3D models with a higher resolution.
Got first results of various skatespots, from single benches to a whole plaza. The data aquisition and the processing of the data has been done within two days.