Fig 1.
The workflow of the proposed approach.
Fig 2.
The model of DIDSON sonar.
Fig 3.
Flow chart of the proposed method.
(i) clustering analysis, (ii) adaptive tensor voting, (iii) minimum spanning tree construction and edge pruning.
Fig 4.
Image detection process results.
a(i∼iii): original images, b(i∼iii): image blocks, c(i∼iii): crack fragments, d(i∼iii): crack probability map, e(i∼iii): final crack curves.
Table 1.
Classification for different crack types.
Fig 5.
Statistical properties of sample characteristic set and test characteristics.
(a) sonar images for 1.8 MHz pattern. (b) sonar images for 1.1 MHz pattern.
Table 2.
The fuzzy rules for BBA values of relative lengths and gray intensity ratios.
Fig 6.
The fusion results from different perspectives alongside a comparison of different frequencies and their fusion.
Table 3.
The BBA values obtained from sonar imagery using different frequencies and the results of the evidence fusion.
Table 4.
The BBA values for different perspectives.
Fig 7.
Image detection results comparing the proposed method and other classical methods.
(a) Original image, (b) Tensor voting, (c) Wasp colony algorithm and (d) The proposed method.
Table 5.
Classification accuracy results for different crack types.