Defect Detection in Printed Circuit Boards Using Semi-Supervised Learning.

Defect inspection is essential in the semiconductor industry to fabricate printed circuit boards (PCBs) with minimum defect rates. However, conventional inspection systems are labor-intensive and time-consuming. In this study, a semi-supervised learning (SSL)-based model called PCB_SS was developed. It was trained using labeled and unlabeled images under two different augmentations. Training and test PCB images were acquired using automatic final vision inspection systems. The PCB_SS model outperformed a completely supervised model trained using only labeled images (PCB_FS). The performance of the PCB_SS model was more robust than that of the PCB_FS model when the number of labeled data is limited or comprises incorrectly labeled data. In an error-resilience test, the proposed PCB_SS model maintained stable accuracy (error increment of less than 0.5%, compared with 4% for PCB_FS) for noisy training data (with as much as 9.0% of the data labeled incorrectly). The proposed model also showed superior performance when comparing machine-learning and deep-learning classifiers. The unlabeled data utilized in the PCB_SS model helped with the generalization of the deep-learning model and improved its performance for PCB defect detection. Thus, the proposed method alleviates the burden of the manual labeling process and provides a rapid and accurate automatic classifier for PCB inspections.

Color compensation of multicolor fish images.

Multicolor fluorescence in situ hybridization (M-FISH) techniques provide color karyotyping that allows simultaneous analysis of numerical and structural abnormalities of whole human chromosomes. Chromosomes are stained combinatorially in M-FISH. By analyzing the intensity combinations of each pixel, all chromosome pixels in an image are classified. Due to the overlap of excitation and emission spectra and the broad sensitivity of image sensors, the obtained images contain crosstalk between the color channels. The crosstalk complicates both visual and automatic image analysis and may eventually affect the classification accuracy in M-FISH. The removal of crosstalk is possible by finding the color compensation matrix, which quantifies the color spillover between channels. However, there exists no simple method of finding the color compensation matrix from multichannel fluorescence images whose specimens are combinatorially hybridized. In this paper, we present a method of calculating the color compensation matrix for multichannel fluorescence images whose specimens are combinatorially stained.

Feature normalization via expectation maximization and unsupervised nonparametric classification for M-FISH chromosome images.

Multicolor fluorescence in situ hybridization (M-FISH) techniques provide color karyotyping that allows simultaneous analysis of numerical and structural abnormalities of whole human chromosomes. Chromosomes are stained combinatorially in M-FISH. By analyzing the intensity combinations of each pixel, all chromosome pixels in an image are classified. Often, the intensity distributions between different images are found to be considerably different and the difference becomes the source of misclassifications of the pixels. Improved pixel classification accuracy is the most important task to ensure the success of the M-FISH technique. In this paper, we introduce a new feature normalization method for M-FISH images that reduces the difference in the feature distributions among different images using the expectation maximization (EM) algorithm. We also introduce a new unsupervised, nonparametric classification method for M-FISH images. The performance of the classifier is as accurate as the maximum-likelihood classifier, whose accuracy also significantly improved after the EM normalization. We would expect that any classifier will likely produce an improved classification accuracy following the EM normalization. Since the developed classification method does not require training data, it is highly convenient when ground truth does not exist. A significant improvement was achieved on the pixel classification accuracy after the new feature normalization. Indeed, the overall pixel classification accuracy improved by 20% after EM normalization.

Maximum-likelihood decomposition of overlapping and touching M-FISH chromosomes using geometry, size and color information.

Since the birth of chromosome analysis by the aid of computers, building a fully automated chromosome analysis system has been the ultimate goal. Along with many other challenges, automating chromosome classification and segmentation has been one of the major challenges especially due to overlapping and touching chromosomes. In this paper we present a novel decomposition method for overlapping and touching chromosomes in M-FISH images. To overcome the limited success of previous decomposition methods that use partial information about a chromosome cluster, we have incorporated more knowledge about the clusters into a maximum-likelihood frame work. The proposed method evaluates multiple hypotheses based on geometric information, pixel classification results, and chromosome sizes, and a hypothesis that has a maximum-likelihood is chosen as the best decomposition of a given cluster. About 90% of accuracy was obtained for two or three chromosome clusters, which consist about 95% of all clusters with two or more chromosomes.

Joint segmentation and classification of M-FISH chromosome images.

Automatic segmentation and classification of M-FISH chromosome images are jointly performed using a six-feature, 25-class maximum-likelihood classifier. Preprocessing of the images including background correction and six-channel color compensation method are introduced. A feature transformation method, spherical coordinate transformation, is introduced. High correct classification results are obtained.

Cyclic capillary electrophoresis.

A strategy is described here for increasing both the resolution and the flexibility of capillary electrophoresis performed in a sieving medium of ungelled polymer. This strategy is based on analysis and, sometimes, re-analysis that is done in several stages of constant-field electrophoresis. Enhancement-stages are between the analysis-stages. An enhancement-stage (i) increases the separation between peaks, while (ii) moving DNA molecules in the reverse direction. An enhancement-stage is based on an electrophoretic ratchet generated by a pulsed electrical field that can be zero-integrated. The ratchet-generating pulses are longer than the field pulses that have previously been used to improve the resolution of DNA molecules. No limit has been found to the resolution enhancement achievable. Apparently, diffusion-induced peak broadening is inhibited and, in some cases, may be reversed by the ratchet. The enhancement-stages are critically dependent on the electrical field-dependence of a plot of electrophoretic mobility as a function of DNA length. To generate the pulsed electrical field, a computer-controlled system with a time resolution of 30 microseconds has been developed. Programming is flexible enough to embed other pulses within ratchet-generating pulses. These other pulses can be either the previously used, shorter field-inversion pulses or high-frequency periodic oscillations previously found to sharpen peaks.