cost function and constraint - icsl-Jeon/image-tracking GitHub Wiki
cost function
distance & tracking cost
easy... just [g-p+view_vector]^2 + (r-r_d)^2
visibility cost
objective: stay way from occluded region in binary matrix
method1
finding maximum box around current azimuth and elevation
pros: easy to formulate, just distance from the center of box => optimization good and stable
cons: 1. finding box in binary matrix ... very inflexible?
2. box approximation to unoccluded region ...
method2
signed distance transform -> interpolation or B-spline surface B-spline:https://en.wikipedia.org/wiki/B-spline
pros: 1. no approximation needed. can exploit cast result as it is. 2. B-spline can be a constraint function ... so better to study. cons: optimization on piece wise? B spline curve
hard constraint
B-spline surface
tutorial on B-spline curve
Ch2 http://www.uio.no/studier/emner/matnat/ifi/INF-MAT5340/v05/undervisningsmateriale/kap2-new.pdf
B spline surface practical
http://www.cgl.uwaterloo.ca/ecdfourq/courses/cs779/project.html
B spline tool box
https://github.com/lasote/multivariate-splines
toolbox guide
https://github.com/bgrimstad/splinter/blob/master/docs/cpp_interface.md
if we make install , default install goes to /usr/local/include for headers and /usr/local/lib for lib files. so link "splinter-static-3-0" to project