YoungJun Roh

Defect Classification Using Bayesian Approach for Tape Substrate Inspection System

Abstract The inspection of ultra-fine pitch patterned tape substrate (TS) requires high resolution optics. In the process of picking out defects at the level of the critical dimension through image processing, however, trivial blemishes formed by dust or micro particles may be detected simultaneously. This leads to unnecessary work on the part of operators reviewing and verifying the additional detected points. To maximize the efficiency of the inspection process, we need to identify and classify the defect candidates whether it is a real pattern defect or simply a trivial blemish by dust. In this article, we propose a Bayesian approach to classify the defective images based on the measures of the image features. The features of the defective region in terms of shape and brightness are obtained from a series of proper image analysis with FFT. Based on the data collected from experiments, we devised a statistic model for classification.

GPU accelerated real time rotation, scale and translation invariant image registration method

This paper presents highly optimized implementation of image registration method that is invariant to rotation scale and translation. Image registration method using FFT works with comparable accuracy as similar methods proposed in the literature, but practical applications seldom use this technique because of high computational requirement. However, this method is highly parallelizable and offloading it to the commodity graphics cards increases its performance drastically. We are proposing the parallel implementation of FFT based registration method on CUDA and OpenCL. Performance analysis of this implementation suggests that the parallel version can be used for real time image registration even for image size up to 2k x 2k. We have achieved significant speed up of up to 345x on NVIDIA GTX 580 using CUDA and up to 116x on AMD HD 6950 using OpenCL. Comparison of our implementation with other GPU based registration method reveals that our implementation performs better compared to other implementations.

A Reflectometry Approach for Rippling Defect Measurement on High Glossy Surface

Locally deformed defects, rippling, on home appliances cause appearance quality issues, and they look enlarged on high glossy surface. The defects need to be detected in the production line, where the product is moving on a conveyor transfer system. Inspection is executed by an expert but it is very subjective and not quantitative so that even quality assurance team cannot manage the defects levels. As a solution, a visual measurement methodology based on one shot reflection Moire is developed to obtain numerical data about surface defects. The system is introduced in this article, and experiment results are shown.

The Novel Approach for Multi-layered Thin Film Inspection Based on 3D Surface Reconstruction

Currently, thin films are widely used for the display product. Furthermore, the protective film is often attached to touch films and cover glasses for their safety. However, relatively few researches have been done for those products and there are many issues when we use current inspection system for multi-layered thin films. In this paper, we propose the novel method to inspect them based on 3d surface reconstruction. We use line profiles of each layer which composes the multi-layered thin film. Also an enhanced light source system is applied to inspect user specified layers by using the patterned retarder. Also we propose the robust defect analysis system by using user specified depth images. We made 3d inspection platform which shows very stable results regarding defects detection among layers. Finally, our study is applied to the commercial system which is used for visual inspection of the film.

High speed Tilted white-light Scanning Interferometry system for package bumps inspection

Tilted white-light Scanning Interferometry(TSI) is oriented from Vertical Scanning white-light interferometry(VSI). TSI is optimized for lateral scanning; its possible that scanning object with only one axis movement while VSI have to move more than two axis to scan a object lager than its Field Of View(FOV). This paper introduces implemented system for package bumps scanning using TSI with our own developed high speed camera.

Phase Map Generation for Phase Shift Moire Using CUDA

Phase shift Moiré is a very popular and one of the most successful techniques for shape measurement of 3-D objects such as PCB (printed circuit board), TFT (thin film transistor), LCD (Liquid crystal display) etc. Various implementations of phase shift moiré are available for improving accuracy and/or speed. Although, these methods contribute a lot in reducing the computation with some compromise in accuracy, there is a lot of scope of improving the performance of these algorithms with increased accuracy, especially when specialized hardware like GPU is available. GPU contains many core or processing elements that can process the same work concurrently resulting in dramatic increase in performance. In this paper, we propose the parallel implementation of the phase shift moiré method on CUDA. A novel method called image stacking method is proposed that can also be used for CUDA implementation of similar algorithms to improve performance. Using this technique, we are able to execute the application 180 times faster compared to the CPU implementation.

A patterned illumination and phase integration imaging method for high-glossy surface evaluation

Visual and optical measurement and evaluation technologies are highly required in quality control of painting and final coating process. We propose a novel approach to evaluate the appearance quality of high glossy surface object. To image the surface roughness, by which the object looks like orange peel, we employ a patterned illumination with phase shifting. The object can be exposed to various illumination conditions according to the phase of pattern lighting. By ac hieving a series of images and analyzing the phase, an image can be synthesized as specific phase selectively. The grade of su rface quality in terms of orange peel is then evaluated based on a specific phase. We have conducted evaluation tests on sample pr oducts, which show the usability of the proposed method.

Development of a Fast and High Resolution X-Ray Imaging Sensor for In-Line Inspection of Tape Substrate

In an automated tape substrate inspections, machine vision is widely adopted for high throughput and cost advantages. However, conventional methods are overly sensitive to foreign particles or have limitations in detecting three-dimensional defects such as top over-etched defect. Foreign particles such as dustsdo not affect the integrity of the final product and are often detected as defects during inspection. To complement vision inspection systems, a prototype fast and fine spatial resolution X-ray imaging sensor has been developed. This image sensor, based on an optoelectronic device – the microchannel plate (MCP), has a spatial resolution of 20 μm and functions at frame rate of 30 fps. X-ray imaging is appropriate as it is virtually transparent to dust particles and provides information regarding the thickness of the copper wire patterns.

A fast and high resolution x-ray imaging sensor for tape substrate inspection

In automated Tape substrate (TS) inspection, machine vision is widely adopted for their high throughput and cost advantages. However, conventional methods are overly sensitive to foreign particles or have limitations in detecting three dimensional defects such as top over-etching. In an attempt to complement vision inspection systems, we proposed utilizing x-ray inspection. To implement x-ray inspection in TS application, we developed a prototype fast and high spatial resolution x-ray imaging sensor which functions at frame rate in excess of 30 fps and has a spatial resolution of 20 µm. In this paper, the development of the sensor and its performance is addressed and the efficiency of the x-ray inspection in detecting top over-etching defects will be shown with experimental studies.