optimization routes, dijkstra algorithm, Biograph,

Hi, when I try to use the dijkstra algorithm to optimization route, always it asks me the segments from the nodes, what happens if I do not know them, because I trying to get the optmization route of an image.

by the way, I do not understand the CMatrix relationship with the real image on the biograph function.

somebody know how to implement that in simulink???

Answer 1

[Walter Roberson]

You need to somehow figure out what the node connections are from the image. The difficulty of doing that will depend on how you create your "nodes" out of the image.

http://www.mathworks.com/matlabcentral/answers/7924-where-can-i-upload-images-and-files-for-use-on-matlab-answers

[Walter Roberson]

Show us an image and point out the parts of it that you want to have considered as being nodes.

(My guess is that you are trying to implement robotic navigation with an array of square tiles, but that is speculation on my part.)

[Image Analyst]

You can find lots of information if you drill down through the links here: http://en.wikipedia.org/wiki/Pathfinding For example, see the right hand panel on the A* page. I've always wanted to try some of the fancy variants but for most cases A* works so well that I've never needed to.

Answer 2

[Image Analyst]

I don't know what a biograph is - apparently according to Google images, it's a celebrity, a theater, or a Bob Dylan album. But if you're trying to find some route through an image, like for example you're following a blood vessel in a radiograph, then the nodes are the pixels. You can do something like find the path from one place to another that has the brightest average gray level (I did this using the A* algorithm as part of my Ph.D.). The Djikstra algorithm has been improved upon many many times so you might want to look into better algorithms. Wikipedia has a huge list of them.

[Walter Roberson]

biograph: http://www.mathworks.com/help/toolbox/bioinfo/ref/biograph.html

[Walter Roberson]

The nodes would be considered to be at the coordinates of the pixels, rather than being the pixels themselves. The background pixels would either not be added to the graph or would be given an infinite cost. The cost of the remaining pixels could be set to be constant, or could be set to be proportional to the pixel value, or you could do contextual image processing in order to determine the cost (e.g., in one area, a slightly darker line might mark a boundary that cannot be passed, and yet due to illumination differences in a different part of the graph, the value associated with that darker area might instead represent areas to be travelled.)

[Image Analyst]

Right, good points. If I recall correctly, in my case, each node had an x (column) position, a y (row, line) position, and a value (gray level). And as you traversed the graph (image), each length was either 1 (up, down, left, right) or sqrt(2) (diagonal travel). And yes the cost function can be specified in a number of ways, some of which could get very complicated, depending on what you want to do.

[Walter Roberson]

Show us an image and point out the parts of it that you want to have considered as being nodes.