“…A clear contrast of the data is obtained by histogram equalization. Histogram equalization is a typical image enhancement method, and its operation is processed by remapping the grayscale levels of the image on the basis of the probability distribution of the input grayscale levels [ 14 ]. To remove the overall noise of the data, we apply a median filtering method.…”
As technology evolves, more components are integrated into printed circuit boards (PCBs) and the PCB layout increases. Because small defects on signal trace can cause significant damage to the system, PCB surface inspection is one of the most important quality control processes. Owing to the limitations of manual inspection, significant efforts have been made to automate the inspection by utilizing high resolution CCD or CMOS sensors. Despite the advanced sensor technology, setting the pass/fail criteria based on small failure samples has always been challenging in traditional machine vision approaches. To overcome these problems, we propose an advanced PCB inspection system based on a skip-connected convolutional autoencoder. The deep autoencoder model was trained to decode the original non-defect images from the defect images. The decoded images were then compared with the input image to identify the defect location. To overcome the small and imbalanced dataset in the early manufacturing stage, we applied appropriate image augmentation to improve the model training performance. The experimental results reveal that a simple unsupervised autoencoder model delivers promising performance, with a detection rate of up to 98% and a false pass rate below 1.7% for the test data, containing 3900 defect and non-defect images.
“…A clear contrast of the data is obtained by histogram equalization. Histogram equalization is a typical image enhancement method, and its operation is processed by remapping the grayscale levels of the image on the basis of the probability distribution of the input grayscale levels [ 14 ]. To remove the overall noise of the data, we apply a median filtering method.…”
As technology evolves, more components are integrated into printed circuit boards (PCBs) and the PCB layout increases. Because small defects on signal trace can cause significant damage to the system, PCB surface inspection is one of the most important quality control processes. Owing to the limitations of manual inspection, significant efforts have been made to automate the inspection by utilizing high resolution CCD or CMOS sensors. Despite the advanced sensor technology, setting the pass/fail criteria based on small failure samples has always been challenging in traditional machine vision approaches. To overcome these problems, we propose an advanced PCB inspection system based on a skip-connected convolutional autoencoder. The deep autoencoder model was trained to decode the original non-defect images from the defect images. The decoded images were then compared with the input image to identify the defect location. To overcome the small and imbalanced dataset in the early manufacturing stage, we applied appropriate image augmentation to improve the model training performance. The experimental results reveal that a simple unsupervised autoencoder model delivers promising performance, with a detection rate of up to 98% and a false pass rate below 1.7% for the test data, containing 3900 defect and non-defect images.
“…With the background subtraction, all of the stationary defects and noise in the images are removed. The linear contrast enhancement and gamma correction (Maragatham & Roomi 2015) are then applied to the images of the two fluid layers separately. Figure 10.( a ) A raw shadowgraphic image showing a magnified view of plume emission near a stable water–FC770 interface (middle dark line).…”
We report a systematic experimental study of the mean temperature profile
$\theta (\delta z)$
and temperature variance profile
$\eta (\delta z)$
across a stable and immiscible liquid–liquid (water–FC770) interface formed in two-layer turbulent Rayleigh–Bénard convection. The measured
$\theta (\delta z)$
and
$\eta (\delta z)$
as a function of distance
$\delta z$
away from the interface for different Rayleigh numbers are found to have the scaling forms
$\theta (\delta z/\lambda )$
and
$\eta (\delta z/\lambda )$
, respectively, with varying thermal boundary layer (BL) thickness
$\lambda$
. By a careful comparison with the simultaneously measured BL profiles near a solid conducting surface, we find that the measured
$\theta (\delta z)$
and
$\eta (\delta z)$
near the liquid interface can be well described by the BL equations for a solid wall, so long as a thermal slip length
$\ell _T$
is introduced to account for the convective heat flux passing through the liquid interface. Direct numerical simulation results further confirm that the turbulent thermal diffusivity
$\kappa _t$
near a stable liquid interface has a complete cubic form,
$\kappa _t(\xi )/\kappa \sim (\xi +\xi _0)^3$
, where
$\kappa$
is the molecular thermal diffusivity of the convecting fluid,
$\xi =\delta z/\lambda$
is the normalized distance away from the liquid interface and
$\xi _0$
is the normalized slip length associated with
$\ell _T$
.
“…In comparison to PSO, Genetic Algorithm, Linear Contrast Stretching, and Histogram Equalization, it was shown that CS performs better. Furthermore, several research have suggested analyzing GIs, with the majority being that meta-heuristics approaches give superior image enhancement when transformation functions are used [16]. Moreover, standardized evaluations for evaluating the efficiency of these techniques are frequently lacking.…”
One of the most important processes in image processing is image enhancement, which aims to enhance image contrast and quality of information. Due to the lack of adequate conventional image enhancement and the challenge of mean shift, intelligence-based image enhancement systems are becoming an essential requirement in image processing. This paper proposes a new approach for enhancing low contrast images utilizing a modified measure and integrating a new Chaotic Crow Search (CCS) and Krill Herd (KH) Optimization-based metaheuristic algorithm. Crow Search Algorithm is a cutting-edge meta-heuristic optimization technique. Chaotic maps are incorporated into the Crow Search Method in this work to improve its global optimization. The new Chaotic Crow Search Algorithm depends on chaotic sequences to replace a random location in the search space and the crow's recognition factor. Based on a new fitness function, Krill Herd optimization is utilized to optimize the tunable parameter. The fitness function requires different primary objective functions that use the image's edge, entropy, grey level co-occurrence matrix (GLCM) contrast, and GLCM energy for increased visual, contrast, and other descriptive information. The results proved that the suggested approach outperforms all-new methods in terms of contrast, edge details, and structural similarity, both subjectively and statistically.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
