Case II: Very Many Tracks
-
In one space (plus time) this tracking problem is formally equivalent to edge detection in vision.
-
In vision, measurements are the discontinuities (differentiations) of color / motion / texture etc. in an image. Edge detection involves linking this basic data into "lines" which will separate regions of image.
-
Solve by neural network method
-
h(x) = 1 if there is an edge
-
= 0 if there is no edge
-
In vision x = (x,y), tracking x = (x,t)
-
h is a field theory formalism, whereas Kalman Filter is a particle formalism
|
