2.2 Image Feature Matching Experiment of RFID System Physical Anti-Collision
45
2.2 Image Feature Matching Experiment of RFID System
Physical Anti-Collision
In the last chapter, we solved the test method of tag group sensitivity and the influence
of different tag distribution on tag group sensitivity. When it is necessary to analyze
the performance of multiple tag groups, we use the means of image acquisition and
processing to confirm the tag position information and complete the evaluation of
tag group performance. This chapter will according to the third chapter of RFID
tags group dynamic testing system, in view of the different key tag location, image
feature matching method to be used for the image obtained by CCD camera system
to extract the characteristics of the key tag, key tags in the complex environment
of effective feature matching and positioning, image feature matching method to
verify the feasibility of performance evaluation of tag clouds. Firstly, SIFT algorithm, SURF algorithm, and ORB algorithm are, respectively, used in this chapter
for image matching test. Three algorithms are compared and analyzed, respectively,
from the execution time of the algorithm, the number of feature points and the number
of correct matching points, and the SURF algorithm which is most suitable for image
feature matching is selected. Then, the algorithm is used to locate the key tags in
the tag group. This study is of great significance to the evaluation and optimization of RFID multi-tag performance and provides an important basis for upgrading
the sensitivity measurement system of tag group to the performance optimization
analysis system.
2.2.1 Image Feature Matching Based on SIFT Algorithm
SIFT algorithm can be called scale-invariant feature conversion algorithm, which
detects key points in the spatial scale of images and extracts invariants of rotation,
scale and position of key points. SIFT algorithm processing speed block, but also
has the advantages of scalability and multidimensionality, can solve the image of
occlusion, noise, light and other interference on the detection target. The steps of SIFT
algorithm can be summarized as generating feature points and describing feature
points.
(1) Generate feature points
The detection of the extremum point is to search all the positions on the scale space,
and then to detect the extremum point with the scale invariable by differential. First,
you need to build a scale space, the SIFT algorithm, using the scale of the image
by the Gaussian kernel space, the main reason is that the Gaussian kernel function
only measures the same kernel function, can retain more the original image, the
characteristics of the Gaussian convolution is expressed as Eqs. (2.9) and (2.10).
45
2.2 Image Feature Matching Experiment of RFID System
Physical Anti-Collision
In the last chapter, we solved the test method of tag group sensitivity and the influence
of different tag distribution on tag group sensitivity. When it is necessary to analyze
the performance of multiple tag groups, we use the means of image acquisition and
processing to confirm the tag position information and complete the evaluation of
tag group performance. This chapter will according to the third chapter of RFID
tags group dynamic testing system, in view of the different key tag location, image
feature matching method to be used for the image obtained by CCD camera system
to extract the characteristics of the key tag, key tags in the complex environment
of effective feature matching and positioning, image feature matching method to
verify the feasibility of performance evaluation of tag clouds. Firstly, SIFT algorithm, SURF algorithm, and ORB algorithm are, respectively, used in this chapter
for image matching test. Three algorithms are compared and analyzed, respectively,
from the execution time of the algorithm, the number of feature points and the number
of correct matching points, and the SURF algorithm which is most suitable for image
feature matching is selected. Then, the algorithm is used to locate the key tags in
the tag group. This study is of great significance to the evaluation and optimization of RFID multi-tag performance and provides an important basis for upgrading
the sensitivity measurement system of tag group to the performance optimization
analysis system.
2.2.1 Image Feature Matching Based on SIFT Algorithm
SIFT algorithm can be called scale-invariant feature conversion algorithm, which
detects key points in the spatial scale of images and extracts invariants of rotation,
scale and position of key points. SIFT algorithm processing speed block, but also
has the advantages of scalability and multidimensionality, can solve the image of
occlusion, noise, light and other interference on the detection target. The steps of SIFT
algorithm can be summarized as generating feature points and describing feature
points.
(1) Generate feature points
The detection of the extremum point is to search all the positions on the scale space,
and then to detect the extremum point with the scale invariable by differential. First,
you need to build a scale space, the SIFT algorithm, using the scale of the image
by the Gaussian kernel space, the main reason is that the Gaussian kernel function
only measures the same kernel function, can retain more the original image, the
characteristics of the Gaussian convolution is expressed as Eqs. (2.9) and (2.10).