Currently doing a software based on database search, principle, I use first calculate the theoretical calculation results of the corresponding curve under different parameter combinations (with a set of two-dimensional data and the curve coordinates), stored in a database, and then I take the curve of the measured curve comparison and database theory, find out the most similar curve, the curve of corresponding parameters is I want results, excuse me what similarity algorithm is used to implement this process?

CodePudding user response:

I think you need to clarify is similar to what you want
Is invalid can give some similar, or only compare the results of fitting curve is similar to
If it is 1, you must first remove noise by algorithm, to calculate the fitting curve
As for the two curves are similar, but also points are based on points, shapes, segmentation, etc.,

I think waiting for you to determine the corresponding algorithm to find good, no matter what your demand is mature algorithms have

CodePudding user response:

Similarity determination algorithm is too much, I only choose some commonly used
A brief Java code of the cosine similarity of

 public class CosineCompare {
/* */cosine similarity algorithm 
Public static void main (String [] args) throws the Exception {
Double [] curve_1={1, 4, 5, 7, 8, 9, 6, 5, 3, 2, 1};//coordinates 
Double [] curve_2={1, 1.2, 8, 20, 9, 8, 6, 5, 3, 2, 1};//coordinates 
Double x=0, y=0, z=0; 
For (int I=0; i X +=curve_1 curve_1 [I] * [I]; 
Y +=curve_2 curve_2 [I] * [I]; 
Z +=curve_1 curve_2 [I] * [I]; 
} 
X=math.h SQRT (x); 
Y=math.h SQRT (y); 
System. The out. Println (z/(x * y)); 
} 
} 

This is a Frechet similarity
https://blog.csdn.net/qq_35570161/article/details/82120424

These are some conventional explanation

https://blog.csdn.net/weixin_39050022/article/details/80732249