Limits...
Automated Internal Classification of Beadless Chinese ZhuJi Fleshwater Pearls based on Optical Coherence Tomography Images

View Article: PubMed Central - PubMed

ABSTRACT

Optical coherence tomography (OCT) has been applied to inspect the internal defect of beadless Chinese ZhuJi fleshwater pearls. A novel fully automated algorithm is proposed to classify between normal and defective sub-layer in nacre layer. Our algorithm utilizes the graph segmentation approach to estimate the up and down boundaries of defect sub-layers from flattened and cropped image, and also proposes the strategy for edge and weight construction in segmentation process. The vertical gradients of boundary pixels are used to make grading decision. The algorithm is tested by typical pearl samples, and achieves 100% classification accuracy. The experiment result shows the feasibility and adaptability of the proposed approach, and proves that the OCT technique combined with proposed algorithm is a potential tool for fast and non-destructive diagnosis of internal structure of beadless pearl.

No MeSH data available.


Result of defective sub-layer segmentation of six typical samples: (a) no defects and cracks (a), and (b–f) contained internal defective sub-layer. Red line: up boundary; Green line: down boundary. Images in left column: flattened images. Images in right column: vertical gradient of flattened images.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC5036028&req=5

f2: Result of defective sub-layer segmentation of six typical samples: (a) no defects and cracks (a), and (b–f) contained internal defective sub-layer. Red line: up boundary; Green line: down boundary. Images in left column: flattened images. Images in right column: vertical gradient of flattened images.

Mentions: To assess the performance of our graph segmentation algorithm and its ability to detect the up and down boundaries of defective layer, we inspected the OCT images of 5 typical samples that were same as samples described in Fig. 1(b–f). Figure 2(b–f) shows the estimated boundaries of these samples. The defect sublayer, in general, was a stripe area, but its shape was various. The advantage of graph method was that it overcomes the influence of shape, bending and jump of the sub-layer which located at boundaries. From both original images of samples with defective sub-layers and their vertical gradient image, it was proved that the graph method could extract the defect area in nacre layer of beadless pearls with different morphological characteristics. The sample showed in Fig. 2(a) was normal pearl without any defects and cracks, but the graph method still estimated the boundaries. Even Fig. 2(a) shows the down boundary is above the up boundary. Generally, the pixel intensity in pearl OCT image decreased from top to bottom. The down boundary (green line in Fig. 2), being defined by light-to-dark change, emerged in the junction between first and second ring round at upper part of image. On the contrary, the up boundary, representing the dark-to-light, appeared at the junction between second and third ring round.


Automated Internal Classification of Beadless Chinese ZhuJi Fleshwater Pearls based on Optical Coherence Tomography Images
Result of defective sub-layer segmentation of six typical samples: (a) no defects and cracks (a), and (b–f) contained internal defective sub-layer. Red line: up boundary; Green line: down boundary. Images in left column: flattened images. Images in right column: vertical gradient of flattened images.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC5036028&req=5

f2: Result of defective sub-layer segmentation of six typical samples: (a) no defects and cracks (a), and (b–f) contained internal defective sub-layer. Red line: up boundary; Green line: down boundary. Images in left column: flattened images. Images in right column: vertical gradient of flattened images.
Mentions: To assess the performance of our graph segmentation algorithm and its ability to detect the up and down boundaries of defective layer, we inspected the OCT images of 5 typical samples that were same as samples described in Fig. 1(b–f). Figure 2(b–f) shows the estimated boundaries of these samples. The defect sublayer, in general, was a stripe area, but its shape was various. The advantage of graph method was that it overcomes the influence of shape, bending and jump of the sub-layer which located at boundaries. From both original images of samples with defective sub-layers and their vertical gradient image, it was proved that the graph method could extract the defect area in nacre layer of beadless pearls with different morphological characteristics. The sample showed in Fig. 2(a) was normal pearl without any defects and cracks, but the graph method still estimated the boundaries. Even Fig. 2(a) shows the down boundary is above the up boundary. Generally, the pixel intensity in pearl OCT image decreased from top to bottom. The down boundary (green line in Fig. 2), being defined by light-to-dark change, emerged in the junction between first and second ring round at upper part of image. On the contrary, the up boundary, representing the dark-to-light, appeared at the junction between second and third ring round.

View Article: PubMed Central - PubMed

ABSTRACT

Optical coherence tomography (OCT) has been applied to inspect the internal defect of beadless Chinese ZhuJi fleshwater pearls. A novel fully automated algorithm is proposed to classify between normal and defective sub-layer in nacre layer. Our algorithm utilizes the graph segmentation approach to estimate the up and down boundaries of defect sub-layers from flattened and cropped image, and also proposes the strategy for edge and weight construction in segmentation process. The vertical gradients of boundary pixels are used to make grading decision. The algorithm is tested by typical pearl samples, and achieves 100% classification accuracy. The experiment result shows the feasibility and adaptability of the proposed approach, and proves that the OCT technique combined with proposed algorithm is a potential tool for fast and non-destructive diagnosis of internal structure of beadless pearl.

No MeSH data available.