Fig 1.
High-voltage cable cross-section images.
(a) and (b) are two different types of high-voltage cable cross-section images, and (c) is the local image of (b).
Fig 2.
Analysis of the cross-sectional image.
(a) High-voltage cable Cross-Section image. (b) The high-voltage cable cross-section gradient diagram.
Fig 3.
Multimodal fusion flowchart.
Fig 4.
Some of the images of this dataset.
Fig 5.
Images of the effect of eliminating scratch-type defects in the cross-section of high-voltage cables.
From left to right, Sample 1 to Sample 5. Fig 5a are original cross-sections of high-voltage cables with scratch-type defects. Fig 5b are the elimination effect images of the improved TV algorithm (Method A) proposed in this paper. Fig 5c are images of the elimination effect of the improved frequency domain filtering method (Method B) proposed in [3].
Table 1.
Average elimination rates of scratch-type defects for Method A and Method B on different sizes of high-voltage cables.
Fig 6.
Conductor number detection effect images.
Fig 6a shows the detection effect after using method A (our), Fig 6b shows the detection effect after using method B, and Fig 6c shows the effect of direct detection.
Table 2.
Average detection rate of the number of conductors for different methods.
Fig 7.
Images of the effect of eliminating scratch-type defects in the cross-section of high-voltage cables.
From left to right, Sample 1 to Sample 5. Fig 7a are cross-sections of high-voltage cables with scratch-type defects. Fig 7b are the eliminative effect images of the improved TV algorithm (Method A) proposed in this paper. Fig 7c are eliminative effect images of the for traditional TV algorithm (Method C).
Table 3.
Comparative data of two methods for eliminating scratch-type defects.
Table 4.
Comparative data on the improvement rate of two methods for eliminating scratch-type defects.