46
2 RFID System Physical Anti-Collision Experimental Verification
Fig. 2.9 Gauss pyramid
model
G(x i , y i , σ ) =
1
2πσ 2 exp
−
(x − x i )
2
+ (y − y i )
2
2σ 2
(2.9)
L(x, y, σ ) = G(x, y, σ ) · I (x, y)
(2.10)
where G(x, y, σ ) represents the Gaussian function, I (x, y) represents the original
image, L(x, y, σ ) represents the convolution operation result of the Gaussian function and the original image, (x, y) represents the pixel position, ∗ represents the
convolution operation, and σ represents the fuzzy coefficient.
After obtaining the Gaussian convolution function, it is necessary to establish
the Gaussian pyramid to represent the scale space. The pyramid model is shown in
Fig. 2.9. The original image is the first group of pyramid models, namely the first
layer at the bottom of the pyramid. Each group is composed of six images of the
same size but with different fuzzy coefficients. The first layer (the bottom image)
of the next group directly samples the penultimate layer of the previous group, thus
ensures the continuity of scale space.
After the completion of the Gaussian pyramid, the construction of the pyramid of
Difference of Gaussian (DOG) needs to continue. The difference Gauss pyramid is
determined by the Gauss pyramid, that is, the second layer of the first group of Gauss
pyramid minus the first layer of the first group of Gauss pyramid, the first layer of
the first group of Gauss difference pyramid is obtained, and so on. Therefore, Gauss
pyramid is the foundation of the pyramid of DOG. In the scale space, the detection
of extreme value points is carried out on the basis of the DOG pyramid, that is, every
pixel in each layer of the image on the Gaussian difference pyramid is compared
with its adjacent 26 pixels, namely, 8 pixels around the same layer and 9 pixels in the
upper and lower layers. In this way, the accuracy of detecting extreme value points
in scale space can be guaranteed. The detection of extreme value points is shown in
Fig. 2.10.
The points obtained above are discrete points and not exact extremum points, so it
is necessary to carry out curve fitting for Gaussian difference function. The Gaussian
2 RFID System Physical Anti-Collision Experimental Verification
Fig. 2.9 Gauss pyramid
model
G(x i , y i , σ ) =
1
2πσ 2 exp
−
(x − x i )
2
+ (y − y i )
2
2σ 2
(2.9)
L(x, y, σ ) = G(x, y, σ ) · I (x, y)
(2.10)
where G(x, y, σ ) represents the Gaussian function, I (x, y) represents the original
image, L(x, y, σ ) represents the convolution operation result of the Gaussian function and the original image, (x, y) represents the pixel position, ∗ represents the
convolution operation, and σ represents the fuzzy coefficient.
After obtaining the Gaussian convolution function, it is necessary to establish
the Gaussian pyramid to represent the scale space. The pyramid model is shown in
Fig. 2.9. The original image is the first group of pyramid models, namely the first
layer at the bottom of the pyramid. Each group is composed of six images of the
same size but with different fuzzy coefficients. The first layer (the bottom image)
of the next group directly samples the penultimate layer of the previous group, thus
ensures the continuity of scale space.
After the completion of the Gaussian pyramid, the construction of the pyramid of
Difference of Gaussian (DOG) needs to continue. The difference Gauss pyramid is
determined by the Gauss pyramid, that is, the second layer of the first group of Gauss
pyramid minus the first layer of the first group of Gauss pyramid, the first layer of
the first group of Gauss difference pyramid is obtained, and so on. Therefore, Gauss
pyramid is the foundation of the pyramid of DOG. In the scale space, the detection
of extreme value points is carried out on the basis of the DOG pyramid, that is, every
pixel in each layer of the image on the Gaussian difference pyramid is compared
with its adjacent 26 pixels, namely, 8 pixels around the same layer and 9 pixels in the
upper and lower layers. In this way, the accuracy of detecting extreme value points
in scale space can be guaranteed. The detection of extreme value points is shown in
Fig. 2.10.
The points obtained above are discrete points and not exact extremum points, so it
is necessary to carry out curve fitting for Gaussian difference function. The Gaussian