6.1 RFID Multi-tag 3D Measurement System
205
image matching method. Finally, DBN is used to model the nonlinear relationship
between the RFID tag group’s 3D positions and corresponding reading distance. The
DBN can predict the reading distance of unknown tag groups and find out the optimal
distribution structure of tag groups corresponding to the maximum reading distance.
The software of the 3D structure prediction system is written in C + + language.
The software consists of four layers: the application interface layer, the system parameter configuration layer, the test protocol layer as well as the data storage and process
layer. The relevant parameters and commands are transmitted between the layers.
The data acquisition electronics of 3D structure prediction systems are mainly
composed of reader, reader antenna, laser rangefinder, and CCD camera. The RFID
reader uses Impinj’s Speedway Revolution R420 ultra high frequency reader. The
reader antenna uses the Larid A9028 far-field antenna. The laser rangefinder uses
Wenglor Company’s X1TA101MHT88 laser rangefinder. The CCD camera lens uses
the Japanese Utron Company’s 2 million pixel level FV0622 industrial lens with a
focal length of 6.5 mm.
6.1.2 Image Process Module
In the RFID tag position measurement system, an image acquisition system is devised
to build a basic measurement platform. In this system, a dual camera collect the multilabel images from two different angles. When the multi-label position is measured,
the perpendicular camera captures the top view of the multi-label, in which the placement on the horizontal plane can be seen. The horizontal camera captures perpendicular multi-image as the turning disc rotates clockwise, where the placement of
the multi-label in the perpendicular direction can be got.
The flowchart of the RFID multi-label localization system is shown in Fig. 6.4.
First, set the speed and height of the turning disc so that all the labels are fully visible
to the camera and the turning disc rotates clockwise at 5 s/rad. The multi-label images
are captured by a dual CCD camera and pre-processed to get clearer images. The
center of the turning disc is marked. The horizontal coordinates of labels can be
calculated by template matching on the perpendicular camera. With the rotation at
a constant speed, the horizontal camera can take the vertical images of the multilabel at different angles, and the location of the multi-label can be computed by
template matching. Finally, SAM converts the coordinates of the image pixel into
multi-label coordinates of the real space. Randomly distribute multi-label again, after
all procedures are completed.
In the image acquisition system, the perpendicular camera firstly acquires a multilabel image in the horizontal direction, as shown in Fig. 6.5a, in which the center
position of the turntable is marked. Second, the horizontal camera acquires multilabel images in the vertical orientation as shown in Fig. 6.5b. The relative motion
between the label and the camera produces motion blur and the noise from various
sources caused by multi-label in optical imaging system are also the main reason
of image degradation. We focus on the multi-label image denoising and deblur to
205
image matching method. Finally, DBN is used to model the nonlinear relationship
between the RFID tag group’s 3D positions and corresponding reading distance. The
DBN can predict the reading distance of unknown tag groups and find out the optimal
distribution structure of tag groups corresponding to the maximum reading distance.
The software of the 3D structure prediction system is written in C + + language.
The software consists of four layers: the application interface layer, the system parameter configuration layer, the test protocol layer as well as the data storage and process
layer. The relevant parameters and commands are transmitted between the layers.
The data acquisition electronics of 3D structure prediction systems are mainly
composed of reader, reader antenna, laser rangefinder, and CCD camera. The RFID
reader uses Impinj’s Speedway Revolution R420 ultra high frequency reader. The
reader antenna uses the Larid A9028 far-field antenna. The laser rangefinder uses
Wenglor Company’s X1TA101MHT88 laser rangefinder. The CCD camera lens uses
the Japanese Utron Company’s 2 million pixel level FV0622 industrial lens with a
focal length of 6.5 mm.
6.1.2 Image Process Module
In the RFID tag position measurement system, an image acquisition system is devised
to build a basic measurement platform. In this system, a dual camera collect the multilabel images from two different angles. When the multi-label position is measured,
the perpendicular camera captures the top view of the multi-label, in which the placement on the horizontal plane can be seen. The horizontal camera captures perpendicular multi-image as the turning disc rotates clockwise, where the placement of
the multi-label in the perpendicular direction can be got.
The flowchart of the RFID multi-label localization system is shown in Fig. 6.4.
First, set the speed and height of the turning disc so that all the labels are fully visible
to the camera and the turning disc rotates clockwise at 5 s/rad. The multi-label images
are captured by a dual CCD camera and pre-processed to get clearer images. The
center of the turning disc is marked. The horizontal coordinates of labels can be
calculated by template matching on the perpendicular camera. With the rotation at
a constant speed, the horizontal camera can take the vertical images of the multilabel at different angles, and the location of the multi-label can be computed by
template matching. Finally, SAM converts the coordinates of the image pixel into
multi-label coordinates of the real space. Randomly distribute multi-label again, after
all procedures are completed.
In the image acquisition system, the perpendicular camera firstly acquires a multilabel image in the horizontal direction, as shown in Fig. 6.5a, in which the center
position of the turntable is marked. Second, the horizontal camera acquires multilabel images in the vertical orientation as shown in Fig. 6.5b. The relative motion
between the label and the camera produces motion blur and the noise from various
sources caused by multi-label in optical imaging system are also the main reason
of image degradation. We focus on the multi-label image denoising and deblur to