230
6 Deep Learning and RFID System Physical Anti-Collision
H i = h ± c i × a
(6.23)
In Eq. (6.23), H i represents the i th tag’s vertical position. h represents the
template’s vertical position. c i is the pixel difference between the template center
point, and the i th tag center point. a is the pixel’s size in the image. When the i th
tag is above the template, Eq. (8) uses the “ + ” sign. When the i th tag is below the
template, Eq. (8) takes the “–” sign.
6.6 Nonlinear Modeling Method Based on DBN
According to the reading distance test module in RRRS, the reading distance of
RFID tag groups is obtained. According to the first three modules in RPMS, 3D
positions of RFID tag groups are obtained. By analyzing the above-obtained data,
it can be concluded that the RFID tags’ 3D positions and corresponding reading
distance have a nonlinear relationship. In this paper, the relationship between the
RFID tags’ 3D positions and corresponding reading distance is built by DBN in the
modeling and prediction module in RPMS. A classical DBN contains an unsupervised learning subpart and a logistic regression layer. Restricted Boltzmann machines
(RBMs) construct the unsupervised learning subpart. The logistic regression layer is
used for prediction.
6.6.1 Restricted Boltzmann Machine
The RBM consists of a visible layer and a hidden layer [34]. Visible variables
and hidden variables are binary variables whose statuses are 0 or 1. The entire
network is a bipartite graph. There are bidirectional and symmetrical connections
between different layers. There is no connection between the neurons within the
same layer. The RBM has excellent feature learning capabilities. The learning of
features is a more abstract expression of data, which is conducive to the extraction
and classification of data.
In Fig. 6.24, v m and h n are the node values in the visible layer and the hidden
layer. W is the weight matrix that connects the visual layer and the hidden layer. b m
and c n are the bias of the visible layer and the hidden layer.
The RBM network is described by using an energy function [32]. The energy
function is used to measure the state of the entire system. The smaller the value
of the energy functions, the more stable the corresponding system is. The RBM’s
energy function is defined as Eq. (6.24).
E(v, h) = −
n
i=1
m
j=1
w i j h i v j −
m
j=1
b j v j −
n
i=1
c i h i
(6.24)
6 Deep Learning and RFID System Physical Anti-Collision
H i = h ± c i × a
(6.23)
In Eq. (6.23), H i represents the i th tag’s vertical position. h represents the
template’s vertical position. c i is the pixel difference between the template center
point, and the i th tag center point. a is the pixel’s size in the image. When the i th
tag is above the template, Eq. (8) uses the “ + ” sign. When the i th tag is below the
template, Eq. (8) takes the “–” sign.
6.6 Nonlinear Modeling Method Based on DBN
According to the reading distance test module in RRRS, the reading distance of
RFID tag groups is obtained. According to the first three modules in RPMS, 3D
positions of RFID tag groups are obtained. By analyzing the above-obtained data,
it can be concluded that the RFID tags’ 3D positions and corresponding reading
distance have a nonlinear relationship. In this paper, the relationship between the
RFID tags’ 3D positions and corresponding reading distance is built by DBN in the
modeling and prediction module in RPMS. A classical DBN contains an unsupervised learning subpart and a logistic regression layer. Restricted Boltzmann machines
(RBMs) construct the unsupervised learning subpart. The logistic regression layer is
used for prediction.
6.6.1 Restricted Boltzmann Machine
The RBM consists of a visible layer and a hidden layer [34]. Visible variables
and hidden variables are binary variables whose statuses are 0 or 1. The entire
network is a bipartite graph. There are bidirectional and symmetrical connections
between different layers. There is no connection between the neurons within the
same layer. The RBM has excellent feature learning capabilities. The learning of
features is a more abstract expression of data, which is conducive to the extraction
and classification of data.
In Fig. 6.24, v m and h n are the node values in the visible layer and the hidden
layer. W is the weight matrix that connects the visual layer and the hidden layer. b m
and c n are the bias of the visible layer and the hidden layer.
The RBM network is described by using an energy function [32]. The energy
function is used to measure the state of the entire system. The smaller the value
of the energy functions, the more stable the corresponding system is. The RBM’s
energy function is defined as Eq. (6.24).
E(v, h) = −
n
i=1
m
j=1
w i j h i v j −
m
j=1
b j v j −
n
i=1
c i h i
(6.24)