Inspired by caving friend Nathan Williams photos of this technique I decided to try to duplicate his results and then write some great software. The idea is to make profiles of cave tunnels known as cross sections very easily and accurately. Cross sections are commonly sketched by a cave mapper by eye with a very rough scale. Sometimes the passage is measured in height and width with a tape.
Here we use a motorized laser level and a DSLR camera to try to construct profiles. After seeing Nathan's photos I got the laser level from Harbor Freight Tools (~$60) and used my Nikon D40X in a local Arkansas cave.
Today I just did a quick test about 100 ft. into the passage. Below is a picture looking toward the level with flash so the tunnel profile can be seen. Then I did a 20 second exposure with the level running and all lights off. There was a small amount of light from the entrance, but negligible.
I then read the image into python, remove tripod reflections by subtracting the average of the blue and green channels from the red and then inverting the resulting monochrome image. The result is seen below:
The big thing I need is the software to then produce a set of points that describe the profile so I can implement routines to compute area and make a pseudo 3-D model of the cave by stacking many closely spaced profiles. I also tested the scale of the image by counting how many pixels wide the level appears and then determined the pixels/cm count to get the size of the tunnel. This process will be improved and automated as the software develops.
I'm open to suggestions from cavers and numerical methods folks. I have a contouring algorithm (Moore-Neighbor Tracing) coded, but it doesn't handle the breaks in the profile. Any ideas on making it continuous and possibly minor smoothing? I plan to build a "T" shape device with 4 dim LEDs to provide a larger scale target.