The Doctoral thesis demonstrates the design of reliable industrial visual systems. The special emphasis is dedicated to the detection of defects on webs in industrial applications based on line-scan cameras. This system makes possible detection and classification of defects originating during the real production conditions. This work covers a theoretical study of a visual system for the defect detection on endless bands as well as of appropriate lighting and the scene arrangement. Further to that have been selected, adjusted and designed key components of hardware. Following the design and optimization of algorithms a system prototype had been installed on non-woven textiles production line. Eight visual systems implemented into real-life industrial conditions based on this prototype

The field of computer vision offers tools that allow machines to derive meaningful infor-mation from video and images and consequently make decisions based on visual inputs. In the paper industry, implementation of machine vision (MV) can be used to automate and speed up processes that require visual inspection, particularly certain segments of quality control – one such application being detection and measurement of paper roll core eccentricity. Core eccentricity is a roll build error in which the roll core is offset from the geometric roll center, potentially causing runnability issues. This particular project aims to improve the detection of paper roll core eccentricity at the Mondi Dynäs integrated pulp and paper mill through creation, calibration and evaluation of a machine-vision-based tool for on-line core eccentricity measurement. The tool utilizes the Hough Transform (HT), since HT is a simple yet fast and robust algorithm when it comes to identification of basic shapes such as lines and circles. The proposed solution was evaluated in two ways; firstly by determining at what level of accuracy the measurements could be provided, accounting for how well the solution deals with correction of systematic error caused by environmental factors, and secondly by analyzing how well characteristic roll features could be accurately identified in large sets of data, necessary to consistently perform measurements. The evaluation of the proposed solution showed a 99.9% detection rate for characteristic paper roll features, and a 98.1% detection rate of laser lines used for correction of position and orientation induced error. Assessment of the measurement accuracy following successful detection was on par with the current optical measurement method, and the proposed solution was able to classify distinctive features with a 96.8% accuracy. Lastly, several improvement actions to address faulty detection were identified, and factors to be considered for future installment were highlighted.

Manual investigation of railway infrastructure is a labor-intensive and time-consuming task, and automating it has become a high priority for railway operators to reduce unexpected infrastructure expenditure. In this thesis, we propose a new image classification approach for classifying defect and non-defective rail joints in image data, based on previous fault detection algorithms using object detection. The rail joints model is to our knowledge a world first, with the vast majority of research into applying computer vision for rail defect detection focusing mainly on the rail tracks and sleepers. Our new image classification models are based on the widely popular Inception ResNet V2, which we fine-tune and compare against a counterpart trained using self-supervision. Additional comparisons are performed against the Faster R-CNN object detector that has had successes with rail tracks and sleepers at the Swiss Federal Railways, as well as against the novel transformer-based DETR architecture. The research has used an in-house object detection annotated dataset from the Swiss railways, recorded in the context of predictive rail maintanance, with rail joints labeled as either defective, or non-defective. Our proposed image classification approach, using either a pre-trained and then fine-tuned, or self-supervised CNN, uses the bounding boxes in a dataset originally intended for object detection, to perform an expanded crop of the images around the rail joint before feeding it to the neural network. Our new image classification approach significantly outperforms object detection neural networks for the task of classifying defective and non-defective rail joints, albeit with the requirement that the rail joint has to be identified prior to classification. Furthermore, our results suggest that the trained models classify defective joints in the test set more consistently than human rail inspectors. The results show that our proposed method can achieve practical performance on unseen data, and can practically be applied to real-life defect detection with high precision and recall, notably on the railways operated by Swiss Federal Railways, SBB CFF FFS. / Manuell inspektion av järnvägsinfrastruktur är en tids- och arbetskrävande uppgift, och automatisering av inspektionerna har på senare tid blivit mer prioriterat av järnvägsoperatörer i syfte att minska oväntade utgifter som uppkommer till följd av undermålig infrastruktur. I det här examensarbetet presenterar vi en ny bildklassificeringsmetod för att klassificera defekta och icke-defekta järnvägsskarvar i bilder tagna från diagnostiska tåg. Modelleringen av järnvägsskarvar som vi har utfört är till vår kännedom något som aldrig gjorts förut för järnvägsinfrastruktur, då majoriteten av forskning inom datorseende för inspektion av järnvägsinfrastruktur historiskt mest har fokuserat på räls och sliprar. Den nya bildklassificeringsmodellen som vi har utvecklat använder den populära arkitekturen Inception ResNet V2, som vi finjusterar och jämför med ett dito som har tränats med självövervakad inlärning. Vidare jämförelser görs mot objektigenkänningsmetoden Faster R-CNN som fungerat väl för sliprar på den schweiziska järnvägen, samt mot den nya transformer-baserade arkitekturen DETR. Forskningen har använt ett dataset annoterat för objektigenkänning från den schweiziska järnvägen, med järnvägsskarvar märkta som defekta, eller icke-defekta. Vår föreslagna bildklassificeringsmetod, med antingen en förtränad och sedan finjusterad CNN, eller en CNN tränad med sjävövervakad inlärning, använder de annoterade boxarna från datasetet för att beskära bilderna runt skarvarna, och sedan klassificera dem. Vår nya metod baserad på bildklassificering presterar väsentligt bättre än neurala nätverk för objektigenkänning, dock med kravet att järnvägsskarven måste ha identifierats i bilden före klassificering. Vidare visar våra resultat att de tränade bildklassificeringsmodellerna klassificerar defekta skarvar i test-setet mer konsekvent än mänskliga järnvägsinspektörer. Resultaten visar att vår nya metod kan användas praktiskt för att upptäcka defekter i verkligheten, med hög precision och recall i data som inte setts under träningen. Specifikt visar vi att de nya modellerna är praktiskt användbara för järnvägen som drivs av Schweiziska Federala Järnvägarna, SBB CFF FFS.

Quality inspection is an essential part of almost any industrial production line. However, designing customized solutions for defect detection for every product can be costlyfor the production line. This is especially the case for short-series production, where theproduction time is limited. That is because collecting and manually annotating the training data takes time. Therefore, a possible method for defect detection using only synthetictraining data focused on geometrical defects is proposed in this thesis work. The methodis partially inspired by previous related work. The proposed method makes use of aninstance segmentation model and pose-estimator. However, this thesis work focuses onthe instance segmentation part while using a pre-trained pose-estimator for demonstrationpurposes. The synthetic data was automatically generated using different data augmentation techniques from a 3D model of a given object. Moreover, Mask R-CNN was primarilyused as the instance segmentation model and was compared with a rival model, HTC. Thetrials show promising results in developing a trainable general-purpose defect detectionpipeline using only synthetic data

No / Advances in micromoulding technology are now allowing mass production of complex, three-dimensional functional products having sub-milligram masses and carefully tailored surface finishes. In order to create a viable manufacturing process for these components, accurate process monitoring and product evaluation are essential in order to highlight process problems and production of substandard parts. The present study describes work implementing a suite of sensors on a commercial micromoulding machine for detailed process interrogation. Evaluation of demoulded products is performed with a single camera based system combined with custom software to allow for three-dimensional characterisation of products during the process.

Monitoring the mechanical integrity of critical structures is extremely important, as mechanical defects can potentially have adverse impacts on their safe operability throughout their service life. Structural defects can be detected by using active structural health monitoring (SHM) approaches, in which a given structure is excited with harmonic mechanical waves generated by actuators. The response of the structure is then collected using sensor(s) and is analyzed for possible defects, with various active SHM approaches available for analyzing the response of a structure to single- or multi-frequency harmonic excitations. In order to identify the appropriate excitation frequency, however, the majority of such methods require a priori knowledge of the characteristics of the defects under consideration. This makes the whole enterprise of detecting structural defects logically circular, as there is usually limited a priori information about the characteristics and the locations of defects that are yet to be detected. Furthermore, the majority of SHM techniques rely on sensors for response collection, with the very same sensors also prone to structural damage. The Surface Response to Excitation (SuRE) method is a broadband frequency method that has high sensitivity to different types of defects, but it requires a baseline. In this study, initially, theoretical justification was provided for the validity of the SuRE method and it was implemented for detection of internal and external defects in pipes. Then, the Comprehensive Heterodyne Effect Based Inspection (CHEBI) method was developed based on the SuRE method to eliminate the need for any baseline. Unlike traditional approaches, the CHEBI method requires no a priori knowledge of defect characteristics for the selection of the excitation frequency. In addition, the proposed heterodyne effect-based approach constitutes the very first sensor-less smart monitoring technique, in which the emergence of mechanical defect(s) triggers an audible alarm in the structure with the defect. Finally, a novel compact phased array (CPA) method was developed for locating defects using only three transducers. The CPA approach provides an image of most probable defected areas in the structure in three steps. The techniques developed in this study were used to detect and/or locate different types of mechanical damages in structures with various geometries.