c++ - OpenCV logo recognition based ICPR 2010 article -

i'm looking on article:

li, zhe, matthias schulte-austum, , martin neschen. "fast logo detection , recognition in document images." pattern recognition (icpr), 2010 20th international conference on. ieee, 2010.

here can find pdf.

i'm not sure of method. implement in opencv.

can know how should implement using opencv?

edit : in article, stated scale , rotation invariant shape descriptor i'm not sure , how use in opencv. idea?

i trying answer, paper poor on details. clear thing definition of descriptor, lacks accurate description regarding:

• layers
• line profiles
• prototypes
• verification procedure

so, you'll find lot of difficulties while implementing method (i know personal experience :d).

my advice (if it's possible) find paper detailed description of method can implement. or find source code of method (some paper cite link code, or find university research teams have link code used).

regarding computation of descriptor (that highlighted in paper):

for every connected component, convex hull calculated. figure 2 gives example. black area connected component , dashed line convex hull of it. convex hull, orientation θ using 2th moments, square root of variance of main axis respect orientation σmax, square root of variance on orthogonal direction σmin, maximum length lmax, minimum edge lmin, , square root of size lsize calculated.

you can in opencv doing:

• connected componets -> findcontours
• convex hull -> convexhull
• moments -> moments
• all others -> need work points on contours, or vertices of convex hull.

my advice still find better described method. not papers @ icpr good, , 1 of that!