This research is to develop an auto optical inspection system for surface mount device light emitting diodes. The principal purpose is to inspect SMD LED for 2D defects which are mixed-material and resin-tearing and for3D defect which is tombstone. In terms with mixed-material inspection, using the count of gradient operator to recognize LED chip. The false alarm rate is 4.29% and misdetection rate is 7.19%. It successfully detects defects with accuracy up to 94.24%. The average computation time is 12.97 ms. In terms of resin-tearing inspection, the research uses the gray scale correlation for SMD LED image registration. The false alarm rate is 5.15% and misdetection rate is 11.34%. The accuracy is up to 91.75%. The average computation time is 10.95 ms. 3D defect continues to use 2D view finder. The advantage of this structure is simple and cost-saving. The investigation which is inspected by the 3D system, comparing with real situation, the average measurement deviation is 4.51%. The average computation time is 8.05 ms. This propose of this system is not only to inspect 2D quickly, precisely and steady, but also to inspect 3D flaws which is hard to detect, and make the wole detective system more artificially-intelligent.

This master's thesis deals with proposal and realization of electromechanical positional system for automatic optical inspection PCB bigger proportions and solution of automatic optical inspection. Problems are dispersed into of several prime area namely on mechanical part, driving hardware and software part. In every part is performed analysis problem and choice optimal solution.

Reverse engineering of hardware IP block is a common practice for competitive purposes in the semiconductor industry. What is done with the information gathered is the deciding legal factor. Once this information gets into the hands of an attacker, it can be used to manufacture exact clones of the hardware device. In an attempt to prevent the illegal copies of the IP block from flooding the market, layout-level obfuscation based on switchable dopant is suggested for the hardware design. This approach can be integrated into the design and manufacturing flow using an obfuscation tool (ObfusTool) to obfuscate the functionality of the IP core. The ObfusTool is developed in a way to be flexible and adapt to different standard cell libraries and designs. It enables easy and accurate evaluation of the area, power and delay v/s obfuscation trades-offs across different design approaches for hardware obfuscation. The ObfusTool is linked to an obfuscation standard cell library which is based on a prototype design created with Obfuscells and 4-input NAND gate. The Obfuscell is a standard cell which is created with switchable functionality based on the assigned dopant configurations. The Obfuscell is combined with other logic gates to form a standard cell library, which can replace any number of existing gates in the IP block without altering it's functionality. A total of 160 different gates are realized using permutated combinations starting with 26 unique gate functions. This design library provide a high level of obfuscation in terms of the number of combinations an adversary has to go through increase to 2 2000 approximately based on the design under consideration. The connectivity of the design has been ignored by previous approaches, which we have addressed in this thesis. The connectivity of a design leaks important information related to inputs and outputs of a gate. We extend the basic idea of dopant-based hardware obfuscation by introducing "dummy wires". The addition of dummy wires not only obfuscates the functionality of the design but also it's connectivity. This greatly reduces the information leakage and complexity of the design increases. To an attacker the whole design appears as one big 'blob'.This also curbs the attempts of brute force attacks. The introduced obfuscation comes at a cost of area and power overhead on an average 5x, which varies across different design libraries.

Elektronikos gamybos rinkoje produkcijos kokybei užtikrinti buvo pasikliaujama žmogaus vizualia apžiūra ir elektroniniais testais. Pasirodžius personaliniams kompiuteriams, gamybos pramonėje pradėta naudoti „kompiuterinė rega”. Procesas, kuriame naudojamas optinis vaizdo sensorius, pvz.: kamera, apibūdinamas kaip automatinė optinė kontrolė – AOK (angl. Automated Optical Inspection - AOI). Darbo tikslas - sukurti spausdintinio montažo plokščių kokybės įvertinimo sistemą panaudojant kompiuterinę regą, išrinkti spausdintinio montažo plokščių kokybės įvertinimui efektyviausius vaizdo apdorojimo algoritmus ir jų parametrus bei įvertinti jų efektyvumą. Taip pat nagrinėjama aplinkos veiksnių (apšvietimo, naudojamos techninės įrangos ir kt.) įtaka sistemos darbo efektyvumui. / Historically the electronics manufacturing market has relied on a combination of human visual inspection and electrical test methods to ensure product quality. With the advent of the personal computer, the use of "machine vision" in industrial applications gradually became more common. Process where optical sensors (i.e., cameras) are used to make specific pass/fail decisions is usually described as Automated Optical Inspection or AOI. There are discussed problems of designing AOI system in this paper. Main goal is to select most efficient image analysis algorithm and to study other parameters that have impact for designing reliable AOI system.

This work deals with the technical possibilities for automated optical inspection and the arrangements for monitoring the implementation of established elements. There are used methods of detection elements, including advanced algorithm processing. With the described methods was created program and each method was tested. The aim was on the clear description of the problem, the optimal design and processing program with objective results with the lowest number of false detection.

Object inspection sensor is a device that is mainly used in the automation of the production process. The Diploma Thesis deals with the familiarization of optical inspection systems, survey and comparison of companies of these systems in the Czech Republic. ELLA-CS company and dafault state output control products in her workplace is presented in the Diploma Thesis. The next part concerns the design of the inspection system for the company ELLA-CS, which is designed on the platform Raspberry. The thesis also deals with the implementation of the inspection system, which describes the hardware and software of this system. The last part is intended to discuss the results and suggestions for improving the inspection.

Convolutional Neural Network (CNN) has been established as a powerful toolto automate various computer vision tasks without requiring any aprioriknowledge. Printed Circuit Board (PCB) manufacturers want to improve theirproduct quality by employing vision based automatic optical inspection (AOI)systems at PCB assembly manufacturing. An AOI system employs classiccomputer vision and image processing techniques to detect variousmanufacturing faults in a PCB assembly. Recently, CNN has been usedsuccessfully at various stages of automatic optical inspection. However, nonehas used 2D image of PCB assembly directly as input to a CNN. Currently, allavailable systems are specific to a PCB assembly and require a lot ofpreprocessing steps or a complex illumination system to improve theaccuracy. This master thesis attempts to design an effective soldering faultdetection system using CNN applied on image of a PCB assembly, withRaspberry Pi PCB assembly as the case in point.Soldering faults detection is considered as equivalent of object detectionprocess. YOLO (short for: “You Only Look Once”) is state-of-the-art fast objectdetection CNN. Although, it is designed for object detection in images frompublicly available datasets, we are using YOLO as a benchmark to define theperformance metrics for the proposed CNN. Besides accuracy, theeffectiveness of a trained CNN also depends on memory requirements andinference time. Accuracy of a CNN increases by adding a convolutional layer atthe expense of increased memory requirement and inference time. Theprediction layer of proposed CNN is inspired by the YOLO algorithm while thefeature extraction layer is customized to our application and is a combinationof classical CNN components with residual connection, inception module andbottleneck layer.Experimental results show that state-of-the-art object detection algorithmsare not efficient when used on a new and different dataset for object detection.Our proposed CNN detection algorithm predicts more accurately than YOLOalgorithm with an increase in average precision of 3.0%, is less complexrequiring 50% lesser number of parameters, and infers in half the time takenby YOLO. The experimental results also show that CNN can be an effectivemean of performing AOI (given there is plenty of dataset available for trainingthe CNN). / Convolutional Neural Network (CNN) har etablerats som ett kraftfullt verktygför att automatisera olika datorvisionsuppgifter utan att kräva någon apriorikunskap. Printed Circuit Board (PCB) tillverkare vill förbättra sinproduktkvalitet genom att använda visionbaserade automatiska optiskainspektionssystem (AOI) vid PCB-monteringstillverkning. Ett AOI-systemanvänder klassiska datorvisions- och bildbehandlingstekniker för att upptäckaolika tillverkningsfel i en PCB-enhet. Nyligen har CNN använts framgångsrikti olika stadier av automatisk optisk inspektion. Ingen har dock använt 2D-bildav PCB-enheten direkt som inmatning till ett CNN. För närvarande är allatillgängliga system specifika för en PCB-enhet och kräver mångaförbehandlingssteg eller ett komplext belysningssystem för att förbättranoggrannheten. Detta examensarbete försöker konstruera ett effektivtlödningsfelsdetekteringssystem med hjälp av CNN applicerat på bild av enPCB-enhet, med Raspberry Pi PCB-enhet som fallet.Detektering av lödningsfel anses vara ekvivalent medobjektdetekteringsprocessen. YOLO (förkortning: “Du ser bara en gång”) ärdet senaste snabba objektdetekteringen CNN. Även om det är utformat förobjektdetektering i bilder från offentligt tillgängliga datasätt, använder viYOLO som ett riktmärke för att definiera prestandametriken för detföreslagna CNN. Förutom noggrannhet beror effektiviteten hos en tränadCNN också på minneskrav och slutningstid. En CNNs noggrannhet ökargenom att lägga till ett invändigt lager på bekostnad av ökat minnesbehov ochinferingstid. Förutsägelseskiktet för föreslaget CNN är inspirerat av YOLOalgoritmenmedan funktionsekstraktionsskiktet anpassas efter vår applikationoch är en kombination av klassiska CNN-komponenter med restanslutning,startmodul och flaskhalsskikt.Experimentella resultat visar att modernaste objektdetekteringsalgoritmerinte är effektiva när de används i ett nytt och annorlunda datasätt förobjektdetektering. Vår föreslagna CNN-detekteringsalgoritm förutsäger merexakt än YOLO-algoritmen med en ökning av den genomsnittliga precisionenpå 3,0%, är mindre komplicerad som kräver 50% mindre antal parametraroch lägger ut under halva tiden som YOLO tar. De experimentella resultatenvisar också att CNN kan vara ett effektivt medel för att utföra AOI (med tankepå att det finns gott om datamängder tillgängliga för utbildning av CNN)

Roll-to-roll manufacturing of micro components based on advanced printing, structuring and lamination of ceramic tapes is rapidly progressing. This large-scale and cost-effective manufacturing process of ceramic micro devices is however prone to hide defects within the visually opaque tape stacks. To achieve a sustainable manufacturing with zero defects in the future, there is an urgent need for reliable inspection systems. The systems to be developed have to perform high-resolution in-process quality control at high speed. Optical coherence tomography (OCT) is a promising technology for detailed in-depth inspection and metrology. Combined with infrared screening of larger areas it can solve the inspection demands in the roll-to-roll ceramic tape processes. In this thesis state-of-art commercial and laboratory OCT systems, operating at the central wavelength of 1.3 µm and 1.7 µm respectively, are evaluated for detecting microchannels, metal prints, defects and delaminations embedded in alumina and zirconia ceramic layers at hundreds of micrometers beneath surfaces. The effect of surface roughness induced scattering and scattering by pores on the probing radiation, is analyzed by experimentally captured and theoretically simulated OCT images of the ceramic samples, while varying surface roughnesses and operating wavelengths. By extending the Monte Carlo simulations of the OCT response to the mid-infrared the optimal operating wavelength is found to be 4 µm for alumina and 2 µm for zirconia. At these wavelengths we predict a sufficient probing depth of about 1 mm and we demonstrate and discuss the effect of rough surfaces on the detectability of embedded boundaries. For high-precision measurement a new and automated 3D image processing algorithm for analysis of volumetric OCT data is developed. We show its capability by measuring the geometric dimensions of embedded structures in ceramic layers, extracting features with irregular shapes and detecting geometric deformations. The method demonstrates its suitability for industrial applications by rapid inspection of manufactured samples with high accuracy and robustness. The new inspection methods we demonstrate are finally analyzed in the context of measurement uncertainty, both in the axial and lateral cases, and reveal that scattering in the sample indeed affects the lateral measurement uncertainty. Two types of image artefacts are found to be present in OCT images due to multiple reflections between neighboring boundaries and inhomogeneity of refractive index. A wavefront aberration is found in the OCT system with a scanning scheme of two galvo mirrors, and it can be corrected using our image processing algorithm. / <p>QC 20140428</p> / Multilayer (FP7-NMP4-2007-214122)