Abstract Currently, structural health monitoring (SHM) systems are not in widespread use for monitoring civil structures because of low defect sensitivity and high cross-sensitivities of most SHM techniques available. The potential drop method (PDM), commonly used in material testing, is a possible method for SHM of large metallic civil structures combining high defect sensitivity and high area monitoring capability. The current state of the art lacks experimental evidence of the applicability to large specimens under the demanding operating conditions of SHM. Here, we investigated the PDM as an SHM system experimentally by analyzing the cross-sensitivity to temperature changes and performed a temperature compensation. We present an optimized method for suppressing the unwanted influence of mechanical loads and increasing the defect sensitivity. The temperature influence was separated from the defect-induced impedance change and effectively suppressed by compensation. Thus, cross-sensitivity does not limit PDM in SHM for large metallic civil structures with temperature compensation. PDM is a promising technique for SHM which could facilitate the widespread use of SHM of conductive civil structures.
{"title":"Investigation of cross-sensitivities of the potential drop method for structural health monitoring of civil structures","authors":"Erik Schneegans, J. Hug, C. Rembe","doi":"10.1515/teme-2023-0024","DOIUrl":"https://doi.org/10.1515/teme-2023-0024","url":null,"abstract":"Abstract Currently, structural health monitoring (SHM) systems are not in widespread use for monitoring civil structures because of low defect sensitivity and high cross-sensitivities of most SHM techniques available. The potential drop method (PDM), commonly used in material testing, is a possible method for SHM of large metallic civil structures combining high defect sensitivity and high area monitoring capability. The current state of the art lacks experimental evidence of the applicability to large specimens under the demanding operating conditions of SHM. Here, we investigated the PDM as an SHM system experimentally by analyzing the cross-sensitivity to temperature changes and performed a temperature compensation. We present an optimized method for suppressing the unwanted influence of mechanical loads and increasing the defect sensitivity. The temperature influence was separated from the defect-induced impedance change and effectively suppressed by compensation. Thus, cross-sensitivity does not limit PDM in SHM for large metallic civil structures with temperature compensation. PDM is a promising technique for SHM which could facilitate the widespread use of SHM of conductive civil structures.","PeriodicalId":56086,"journal":{"name":"Tm-Technisches Messen","volume":null,"pages":null},"PeriodicalIF":1.0,"publicationDate":"2023-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"90820393","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Abstract Elastomers with a percolative network of carbon particles are a frequently studied class of materials for applications requiring high elongation and compliant sensors. For novel applications such as soft robots or smart textiles, these have some advantages over traditional strain gauges. However, their functionality is not fully understood. In this work, such materials are investigated as strain sensors in terms of their dynamic behavior, and their current limitations are demonstrated. It becomes clear that such sensors exhibit a non-monotonic behavior under dynamic loads that differs significantly from that observed in quasi-static tests. Two strategies for improving sensor characteristics are derived, modeled, and experimentally tested using the results and an electro-mechanical network model. First, a melt-spinning process that orients the carbon nanotube particles in the process direction creates different degrees of anisotropy. Second, to generate a local negative transverse contraction, an additional auxetic support structure is used. While the resulting anisotropy is insufficient to improve sensor properties, the auxetic structure can significantly improve strain sensitivity.
{"title":"Properties and special phenomena of strain sensors made of carbon particle-filled elastomers","authors":"J. Mersch, G. Gerlach","doi":"10.1515/teme-2023-0022","DOIUrl":"https://doi.org/10.1515/teme-2023-0022","url":null,"abstract":"Abstract Elastomers with a percolative network of carbon particles are a frequently studied class of materials for applications requiring high elongation and compliant sensors. For novel applications such as soft robots or smart textiles, these have some advantages over traditional strain gauges. However, their functionality is not fully understood. In this work, such materials are investigated as strain sensors in terms of their dynamic behavior, and their current limitations are demonstrated. It becomes clear that such sensors exhibit a non-monotonic behavior under dynamic loads that differs significantly from that observed in quasi-static tests. Two strategies for improving sensor characteristics are derived, modeled, and experimentally tested using the results and an electro-mechanical network model. First, a melt-spinning process that orients the carbon nanotube particles in the process direction creates different degrees of anisotropy. Second, to generate a local negative transverse contraction, an additional auxetic support structure is used. While the resulting anisotropy is insufficient to improve sensor properties, the auxetic structure can significantly improve strain sensitivity.","PeriodicalId":56086,"journal":{"name":"Tm-Technisches Messen","volume":null,"pages":null},"PeriodicalIF":1.0,"publicationDate":"2023-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"78595317","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
L. Bifano, Markus Michel, Max Weidl, A. Fischerauer, G. Fischerauer
Abstract This work deals with the determination of the uncertainty of measurement data, determined by electrical impedance spectroscopy. Four different types of sand were measured impedimetrically in a measuring cell designed as a plate capacitor in a frequency range from 20 Hz to 1 MHz. The measuring cell was filled ten times with each sand and 20 impedance spectra were recorded for each filling. The uncertainty at each frequency was determined from the measurement data. It was found that the measurement data variance with a given measuring-cell filling was negligibly small. However, it increased by a factor of up to 100 when the measuring cell was repeatedly emptied and re-filled with the same material. We propose a way to estimate a continuous approximation of the uncertainty band of the impedance locus in the complex plane from the discrete uncertainties at each frequency. It uses a Support Vector Machine (SVM) to generate a regression curve using the discrete uncertainties. The result of the regression was used to estimate the uncertainties of an average impedance locus. The said machine learning tool can handle large amounts of data, classes, and influencing variables. In this manner, it can help to identify cause-effect relationships. Furthermore, at the end of this work a possibility to estimate a continuous uncertainty band along the impedance locus curve via SVM regression is shown. This is an extension to the common methodology in literature, where the uncertainty is only determined at selected individual points of the impedance spectrum.
{"title":"Machine learning approach for impedance locus uncertainties","authors":"L. Bifano, Markus Michel, Max Weidl, A. Fischerauer, G. Fischerauer","doi":"10.1515/teme-2023-0048","DOIUrl":"https://doi.org/10.1515/teme-2023-0048","url":null,"abstract":"Abstract This work deals with the determination of the uncertainty of measurement data, determined by electrical impedance spectroscopy. Four different types of sand were measured impedimetrically in a measuring cell designed as a plate capacitor in a frequency range from 20 Hz to 1 MHz. The measuring cell was filled ten times with each sand and 20 impedance spectra were recorded for each filling. The uncertainty at each frequency was determined from the measurement data. It was found that the measurement data variance with a given measuring-cell filling was negligibly small. However, it increased by a factor of up to 100 when the measuring cell was repeatedly emptied and re-filled with the same material. We propose a way to estimate a continuous approximation of the uncertainty band of the impedance locus in the complex plane from the discrete uncertainties at each frequency. It uses a Support Vector Machine (SVM) to generate a regression curve using the discrete uncertainties. The result of the regression was used to estimate the uncertainties of an average impedance locus. The said machine learning tool can handle large amounts of data, classes, and influencing variables. In this manner, it can help to identify cause-effect relationships. Furthermore, at the end of this work a possibility to estimate a continuous uncertainty band along the impedance locus curve via SVM regression is shown. This is an extension to the common methodology in literature, where the uncertainty is only determined at selected individual points of the impedance spectrum.","PeriodicalId":56086,"journal":{"name":"Tm-Technisches Messen","volume":null,"pages":null},"PeriodicalIF":1.0,"publicationDate":"2023-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"87430331","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Zusammenfassung Die hochgenaue und zuverlässige Messung von Großbauteilen für Windenergiesysteme auf Groß-Koordinatenmessgeräten (Groß-KMG) ist für deren störungsfreien Einsatz unter höchsten Belastungen unerlässlich. Um der Industrie auf diesem Gebiet Kalibrierungen auf höchstem Niveau sowie weitere messtechnische Unterstützung anbieten zu können, wurde in der Physikalisch-Technischen Bundesanstalt ein Groß-KMG etabliert. In diesem Beitrag werden die jüngsten Entwicklungen zu neuen Ansätzen für die Geometriefehlererfassung für große Messvolumina vorgestellt. Darüber hinaus gibt es Einblick in die Optimierungsstrategie für das neuartige Kalibrierverfahren M3D3 sowie in die Erweiterung des Virtuellen Koordinatenmessgerätes (VCMM) für die simulationsbasierte Messunsicherheitsbestimmung auf Groß-KMG.
{"title":"Rückführung dimensioneller Messungen an Großbauteilen für Windenergiesysteme","authors":"K. Kniel, Daniel Heißelmann, M. Franke","doi":"10.1515/teme-2023-0058","DOIUrl":"https://doi.org/10.1515/teme-2023-0058","url":null,"abstract":"Zusammenfassung Die hochgenaue und zuverlässige Messung von Großbauteilen für Windenergiesysteme auf Groß-Koordinatenmessgeräten (Groß-KMG) ist für deren störungsfreien Einsatz unter höchsten Belastungen unerlässlich. Um der Industrie auf diesem Gebiet Kalibrierungen auf höchstem Niveau sowie weitere messtechnische Unterstützung anbieten zu können, wurde in der Physikalisch-Technischen Bundesanstalt ein Groß-KMG etabliert. In diesem Beitrag werden die jüngsten Entwicklungen zu neuen Ansätzen für die Geometriefehlererfassung für große Messvolumina vorgestellt. Darüber hinaus gibt es Einblick in die Optimierungsstrategie für das neuartige Kalibrierverfahren M3D3 sowie in die Erweiterung des Virtuellen Koordinatenmessgerätes (VCMM) für die simulationsbasierte Messunsicherheitsbestimmung auf Groß-KMG.","PeriodicalId":56086,"journal":{"name":"Tm-Technisches Messen","volume":null,"pages":null},"PeriodicalIF":1.0,"publicationDate":"2023-06-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81205215","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2023-06-21eCollection Date: 2024-03-01DOI: 10.1515/teme-2023-0002
Florian Scheible, Raphael Lamprecht, Casey Schaan, Reinhard Veltrup, Marion Semmler, Alexander Sutor
This paper presents a measurement setup which is able to measure the distribution of small scale pressure on an area of 15.2 mm × 30.4 mm with a sample rate up to 1.2 kHz. It was used to investigate the contact pressures of vocal folds during phonation. This was performed in ex vivo experiments of 11 porcine larynges. The contact pressure at the medial surface and other phonation parameters, as the glottal resistance and the closing velocity of the vocal fold, were measured at different adduction and elongation levels and air flow rates. A statistical analysis was carried out. It could be shown that the contact pressure rises, when the vocal fold is manipulated or when the flow rate is increased.
本文介绍了一种测量装置，它能够测量 15.2 mm × 30.4 mm 面积上的小尺度压力分布，采样率高达 1.2 kHz。它被用于研究发音时声带的接触压力。这是在 11 头猪喉部的体外实验中进行的。在不同的伸展水平和气流速率下，测量了声带内侧表面的接触压力和其他发音参数，如声门阻力和声带闭合速度。结果进行了统计分析。结果表明，当操作声带或增加流速时，接触压力会升高。
{"title":"What kind of phonation causes the strongest vocal fold collision? - A hemi-larynx phonation contact pressure study.","authors":"Florian Scheible, Raphael Lamprecht, Casey Schaan, Reinhard Veltrup, Marion Semmler, Alexander Sutor","doi":"10.1515/teme-2023-0002","DOIUrl":"10.1515/teme-2023-0002","url":null,"abstract":"<p><p>This paper presents a measurement setup which is able to measure the distribution of small scale pressure on an area of 15.2 mm × 30.4 mm with a sample rate up to 1.2 kHz. It was used to investigate the contact pressures of vocal folds during phonation. This was performed in ex vivo experiments of 11 porcine larynges. The contact pressure at the medial surface and other phonation parameters, as the glottal resistance and the closing velocity of the vocal fold, were measured at different adduction and elongation levels and air flow rates. A statistical analysis was carried out. It could be shown that the contact pressure rises, when the vocal fold is manipulated or when the flow rate is increased.</p>","PeriodicalId":56086,"journal":{"name":"Tm-Technisches Messen","volume":null,"pages":null},"PeriodicalIF":1.0,"publicationDate":"2023-06-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10995662/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"72920053","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Abstract Ultrasound is a mechanical wave propagating in tissue which is influenced in its propagation behavior by the locally prevailing acousto-mechanical conditions. By suitable processing of the back-scattered signals received by the ultrasound transducer, tissue parameters such as local bulk modulus, mass density, speed of sound, isotropic scattering coefficient, and also the locally acting tissue absorption can be inferred. A discipline that has received increasing attention in the medical ultrasonic imaging discipline and its scientific publications in recent years is quantitative ultrasound (QUS) which tries to estimate with great accuracy these local acting tissue parameters. In this paper we analyze different algorithms for estimation of high spatial resolution tissue absorption parameters. On the one hand, there is a simple absorption estimator based on the evaluation of the quotient of the power density spectra calculated for different depth regions (spectral-log-difference estimator), which, however, assumes a linearly with frequency increasing absorption, this is contrasted with an estimator which also allows to estimate a polynomial increase of the absorption with frequency (method-of-moments estimator). Since a closed-form solution cannot be given for this, a maximum-likelihood estimator for which there is always an estimate that can be computed numerically efficiently is developed. The results, tissue attenuation, are presented as a color-coded overlay on conventional B-mode ultrasound images showing only morphology.
{"title":"Estimators of tissue absorption parameters power-law prefactor and power-law exponent from medical ultrasonic images","authors":"D. Brandner, B. Zagar","doi":"10.1515/teme-2023-0090","DOIUrl":"https://doi.org/10.1515/teme-2023-0090","url":null,"abstract":"Abstract Ultrasound is a mechanical wave propagating in tissue which is influenced in its propagation behavior by the locally prevailing acousto-mechanical conditions. By suitable processing of the back-scattered signals received by the ultrasound transducer, tissue parameters such as local bulk modulus, mass density, speed of sound, isotropic scattering coefficient, and also the locally acting tissue absorption can be inferred. A discipline that has received increasing attention in the medical ultrasonic imaging discipline and its scientific publications in recent years is quantitative ultrasound (QUS) which tries to estimate with great accuracy these local acting tissue parameters. In this paper we analyze different algorithms for estimation of high spatial resolution tissue absorption parameters. On the one hand, there is a simple absorption estimator based on the evaluation of the quotient of the power density spectra calculated for different depth regions (spectral-log-difference estimator), which, however, assumes a linearly with frequency increasing absorption, this is contrasted with an estimator which also allows to estimate a polynomial increase of the absorption with frequency (method-of-moments estimator). Since a closed-form solution cannot be given for this, a maximum-likelihood estimator for which there is always an estimate that can be computed numerically efficiently is developed. The results, tissue attenuation, are presented as a color-coded overlay on conventional B-mode ultrasound images showing only morphology.","PeriodicalId":56086,"journal":{"name":"Tm-Technisches Messen","volume":null,"pages":null},"PeriodicalIF":1.0,"publicationDate":"2023-06-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81145701","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Christian Kinzig, Guanzhi Feng, Miguel Granero, C. Stiller
Abstract Image-based object detection is a crucial task in autonomous driving. In many cases, objects are not correctly detected and classified if they are only partially visible due to a limited field of view. Also, even if stitched panoramic images are used, errors in object detection can still occur if the seam between individual images is visible. This happens due to vignetting or different exposure, although the images are optimally aligned. In this article, we present a real-time capable and effective method for vignetting compensation and exposure correction. Before runtime, the camera response function is determined and the vignetting model is preliminarily approximated. We obtain the irradiance from the intensity values of incoming images. Then, the vignetting model is applied. Afterwards, the pixels at the seam are used to correct the exposure. Finally, we convert the corrected irradiance back to intensity values. We evaluate our approach by measuring the image stitching accuracy in the overlapping area by the IoU of grayscale histograms and the mean absolute error of intensity values. The metrics are applied both on data recorded with our experimental vehicle and on the publicly available nuScenes dataset. Finally, we demonstrate that our approach runs in real-time on GPU.
{"title":"Real-time vignetting compensation and exposure correction for panoramic images by optimizing irradiance consistency","authors":"Christian Kinzig, Guanzhi Feng, Miguel Granero, C. Stiller","doi":"10.1515/teme-2023-0011","DOIUrl":"https://doi.org/10.1515/teme-2023-0011","url":null,"abstract":"Abstract Image-based object detection is a crucial task in autonomous driving. In many cases, objects are not correctly detected and classified if they are only partially visible due to a limited field of view. Also, even if stitched panoramic images are used, errors in object detection can still occur if the seam between individual images is visible. This happens due to vignetting or different exposure, although the images are optimally aligned. In this article, we present a real-time capable and effective method for vignetting compensation and exposure correction. Before runtime, the camera response function is determined and the vignetting model is preliminarily approximated. We obtain the irradiance from the intensity values of incoming images. Then, the vignetting model is applied. Afterwards, the pixels at the seam are used to correct the exposure. Finally, we convert the corrected irradiance back to intensity values. We evaluate our approach by measuring the image stitching accuracy in the overlapping area by the IoU of grayscale histograms and the mean absolute error of intensity values. The metrics are applied both on data recorded with our experimental vehicle and on the publicly available nuScenes dataset. Finally, we demonstrate that our approach runs in real-time on GPU.","PeriodicalId":56086,"journal":{"name":"Tm-Technisches Messen","volume":null,"pages":null},"PeriodicalIF":1.0,"publicationDate":"2023-06-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88093934","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Abstract Cup anemometers are within the most used wind speed sensors for the wind energy industry. Wind tunnel calibrations under controlled conditions are required, but during operation uncontrolled environmental conditions occur. This is accounted for in the IEC 61400-50-1:2022 international wind measurements standard, which specifies sensor classification based on their response to external conditions, due to the influence parameters turbulence, air temperature, density, and upflow angle. Temperature and density effects are not covered appropriately in the IEC 61400-50-1:2022, since it assumes that air temperature only influences the bearing friction of a cup anemometer. No guidance is provided on evaluating variations in density, which depends on temperature and pressure. To investigate this, two cup anemometers are measured in Deutsche WindGuard’s Climatic Wind Tunnel, where density is changed by varying pressure and temperature independently. The results show that the sensor’s response to temperature can have other effects than an increase in ball bearing friction. Using pressure or temperature to modify density can even cause opposing results. Hence, varying temperatures and pressures independently is crucial to characterize a sensor’s response. The results correspond to cup anemometers, but the methodology is applicable on all sensors.
{"title":"Temperature and pressure effects on the response behavior of anemometers","authors":"Alina Roß, N. Balaresque, A. Fischer","doi":"10.1515/teme-2023-0059","DOIUrl":"https://doi.org/10.1515/teme-2023-0059","url":null,"abstract":"Abstract Cup anemometers are within the most used wind speed sensors for the wind energy industry. Wind tunnel calibrations under controlled conditions are required, but during operation uncontrolled environmental conditions occur. This is accounted for in the IEC 61400-50-1:2022 international wind measurements standard, which specifies sensor classification based on their response to external conditions, due to the influence parameters turbulence, air temperature, density, and upflow angle. Temperature and density effects are not covered appropriately in the IEC 61400-50-1:2022, since it assumes that air temperature only influences the bearing friction of a cup anemometer. No guidance is provided on evaluating variations in density, which depends on temperature and pressure. To investigate this, two cup anemometers are measured in Deutsche WindGuard’s Climatic Wind Tunnel, where density is changed by varying pressure and temperature independently. The results show that the sensor’s response to temperature can have other effects than an increase in ball bearing friction. Using pressure or temperature to modify density can even cause opposing results. Hence, varying temperatures and pressures independently is crucial to characterize a sensor’s response. The results correspond to cup anemometers, but the methodology is applicable on all sensors.","PeriodicalId":56086,"journal":{"name":"Tm-Technisches Messen","volume":null,"pages":null},"PeriodicalIF":1.0,"publicationDate":"2023-06-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"75773093","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Abstract Machine-learning-based (ML) segmentation in the image domain can be utilized for the detection of corrosion on the surface of industrial objects. This research provides a comparison of techniques using convolutional neural networks (CNNs) on the one hand, and random forest (RF) classifiers within RGB and HSV feature spaces on the other hand. CNN-based approaches usually need a large amount of data for training in order for the network to converge and generalize well on new data. Due to the low amount of data provided, we apply a set of methods to increase the generalization ability of the model. These methods can be categorized into data augmentation, selection of larger and smaller models and pretraining strategies like self supervised learning (SSL). The RF classifiers on the other hand are trained per pixel, so that the amount of data is determined by the image size. The object to be tested is a barrel made of metal, from which the image of the coat is used as the training data, and the image of the bottom as test data. We found that a RF classifier in the RGB feature space outperforms the CNNs by seven percentage points regarding the f 1-score of the corrosion class.
{"title":"A comparison of learning-based approaches for the corrosion detection on barrels in industrial applications","authors":"D. Haitz, P. Hübner, M. Ulrich, B. Jutzi","doi":"10.1515/teme-2023-0009","DOIUrl":"https://doi.org/10.1515/teme-2023-0009","url":null,"abstract":"Abstract Machine-learning-based (ML) segmentation in the image domain can be utilized for the detection of corrosion on the surface of industrial objects. This research provides a comparison of techniques using convolutional neural networks (CNNs) on the one hand, and random forest (RF) classifiers within RGB and HSV feature spaces on the other hand. CNN-based approaches usually need a large amount of data for training in order for the network to converge and generalize well on new data. Due to the low amount of data provided, we apply a set of methods to increase the generalization ability of the model. These methods can be categorized into data augmentation, selection of larger and smaller models and pretraining strategies like self supervised learning (SSL). The RF classifiers on the other hand are trained per pixel, so that the amount of data is determined by the image size. The object to be tested is a barrel made of metal, from which the image of the coat is used as the training data, and the image of the bottom as test data. We found that a RF classifier in the RGB feature space outperforms the CNNs by seven percentage points regarding the f 1-score of the corrosion class.","PeriodicalId":56086,"journal":{"name":"Tm-Technisches Messen","volume":null,"pages":null},"PeriodicalIF":1.0,"publicationDate":"2023-06-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"81845342","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：