Information Technology Reference
In-Depth Information
previously, which are regarded as the dictionary of the feature match, i.e. the
number of the dictionary column is 360. Each column represents a projection
unit of a certain angle. This dictionary is used both in the MCA and the pro-
posed method. It can be found from Fig.5 that the accuracy of our method is
close to that of the MCA and better than that of the OCM.
(a)
(b)
(c)
(d)
(e)
Fig. 5. Sample image of a cap matched against the template using three methods.
(a) Template image. (b) Sample image with defect. (c) Rotation matching result using
OCM. (d) Rotation matching result using MCA. (e) Rotation matching result using
our method.
Secondly, in order to verify the running time we select three sample bottle
caps surface images to run 10 times to obtain the average time. The results are
presented in Table 1. Our method is 23 times faster than the OCM and nearly
2 times faster than the MCA. The reason that the OCM is slow may be caused
by many invert tangent function and vast loops process. Our method is superior
to the MCA because we improved algorithm of the finding α t .
Tabl e 1. Result of the average run-time (second) for the three methods
Sample Cap
OCM(second) MCA(second) Our Method(second)
PEARL RIVER 1
0.448263
0.038351
0.019446
PEARL RIVER 2
0.446221
0.038062
0.019001
PEARL RIVER 3
0.450014
0.040433
0.021135
6Con lu on
A new method base on circular region projection histogram and sparse represen-
tation for detecting the defect of caps surface is presented. The method employs
the projection of the template on different orientation as the template dictio-
nary and the circular point projection structure the diagonal matrix as the defect
 
Search WWH ::




Custom Search