Correction of Visual Perception Based on Neuro-Fuzzy Learning for the Humanoid Robot TEO

. 2018 Mar 25;18(4):972. doi: 10.3390/s18040972

Algorithm 2 Data acquisition by using a Neuro-Fuzzy filter. (Figure 6 bottom).

Input: capture image

Output: extract

{\hat{α}}_{r e a l}, C o G (X, Y)

values

Require: camera ON

1:
Capture (Img)
2:
Binarize (Img) //Threshold of the object to be detected
3:
if (Object_detected (Img) == true) then
4:
//Extraction values from the original frame
5:
Bottle = Get_Blobs(Img)
6:
( $X_{C o G}, Y_{C o G}$ ) = Get_CoG(Bottle)
7:
$α$ = Get_Angle_Inclination(Bottle)
8:
end if
9:
10:
//Applying of the Fuzzy filter previously trained
11:
while (i < n) do
12:
//Search of similarities within the dataset
13:
if (( $L i_{n} < X_{C o G} < L s_{n}$ ) and ( $R i_{n} < Y_{C o G} < R s_{n}$ ) and ( $P i_{n} < α < P s_{n}$ )) then
14:
${\hat{α}}_{r e a l} = A n g l e_{e r r o r} - α$
15:
return
16:
end if
17:
i++
18:
end while
19:
return $X_{C o G}, Y_{C o G}, {\hat{α}}_{r e a l}$ // ${\hat{α}}_{r e a l}$ is the real inclination angle