Algorithm for following the path of ridges in a three-dimensional image

Question

I'm trying to find an algorithm (or algorithm ideas) for tracking a ridge in a three-dimensional image derived from a digital elevation model (DEM). I managed to get a very basic program that simply iterates over each line of the image denoting the ridge line, wherever it detects a large change in aspect (from 180 ° to> 180 degrees).

However, the lines they produce are not shiny, often there are gaps and various strange artifacts. I hope to try to expand this using some kind of algorithm that follows the lines of the ridges, thereby creating lines that are complete (i.e. without spaces) and more accurate.

Many people mentioned me snake algorithms, but they don't seem to be exactly what I'm looking for. I also searched a lot for path finding algorithms, but again, they don't seem to be quite right.

Does anyone have any suggestions for types or algorithms or specific algorithms that I should look at?

Update: . I was asked to add more details about the exact area to which I will apply this. It works with trellised data on the height of sand dunes. I am trying to extract ridges if these sand dunes, which are similar to the boundaries between drainage basins, but can be much more complicated (for example, there may be several sand dunes very close to each other with the gradual merging of the ridges)

Answer 1

. , . , :

for each face in the mesh
   compute 1/curvature
   if abs(1/curvature) != zeroTolerance 
     flag face as ridge
   else
     continue

(zeroTolerance - , , , , 0,003 ..)

Meshlab . , , .

Answer 2

, . , ( , , ).

startPoint = highest point in DEM (or on ridge)
curPoint = startPoint;
line += curPoint;
Loop
    curPoint = highest point adjacent to curPoint not in line; // (Don't backtrack)
    line += point;
Repeat

, .

: "" .

Answer 3

http://en.wikipedia.org/wiki/Ridge_detection

Answer 4

You can process the height, just like the grayscale color, and then use the two-sided recognition filter. There are many methods for edge recognition. The best will depend on your specific needs.