Implementing machine learning: chances and challenges

Finding and implementing a suitable machine learning (ML) solution to a task at hand has several facets. The technical side of ML has widely been discussed in detail, see, e. g., (Heizmann, M., A. Braun, M. Hüttel, C. Klüver, E. Marquardt, M. Overdick and M. Ulrich. 2020. Artificial Intelligence with Neural Networks in Optical Measurement and Inspection Systems. at – Automatisierungstechnik 68(6): 477–487). This contribution focusses on the industrial implementation issues of ML projects, particularly for machine vision (MV) tasks. Especially in small and medium-sized enterprises (SMEs), resources cannot be activated at will in order to use a new technology like ML. We take this into account by, on the one hand, helping to realistically evaluate the opportunities and challenges involved in implementing ML projects for a given task. On the other hand, we consider not only technical aspects, but also organizational, social and customer-related ones. It is discussed which know-how a company itself has to bring into an ML project and which tasks can also be performed by service providers. Here, it becomes clear that ML techniques can be used at different levels of detail. The question of “make or buy” is therefore also an entrepreneurial one when introducing ML into one’s own products and processes, and must be answered with a view to one’s own possibilities and structures.

Radioactive Contamination Countermeasures, Food Inspection Systems, and the Issue of Reputational Damage in the Early Stages of the Nuclear Disaster in Fukushima Prefecture

This paper overviews the achievements and challenges of radioactive contamination countermeasures, food inspection systems, and reputational damage to agricultural products in Fukushima Prefecture during the early stages of the Great East Japan Earthquake and nuclear disaster. It outlines the effectiveness of early countermeasures such as absorption control measures and soil decontamination, and observes how efforts aimed at revitalizing afflicted areas were initiated and advanced primarily through the leadership of residents and agricultural producers. Furthermore, it examines food inspection systems such as the “all-bag-all-volume” testing system for rice that was implemented in Fukushima, and suggests that a failure to extend such countermeasures to outside of Fukushima Prefecture was a contributing factor to the ongoing issue of reputational damage and consumer reluctance to purchase products from the area. Lastly, the paper categorizes early consumer trends in four groups based on differing perceptions of risk and safety, and concludes that dealing with reputational damage should entail creating maps of radioactive material distribution, and also building a rational inspection system that allows consumers to objectively identify the safety of agricultural products.

Coded excitation for low-power guided ultrasonic wave inspection in safety-critical industries: case studies

High-power ultrasonic non-destructive evaluation (NDE) poses significant threats to intrinsic safety. It may lead to hazards in critical industrial applications, especially in oil & gas refineries, high-energy material technologies and the aerospace and aviation industries. Typically, industries employ various certifications and undertake several safety protocols to suppress the likelihood of industrial hazards. In order to satisfy safety standards for operating high-power equipment close to potential explosives and inflammable substances, industries direct large sums of investment into making these inspection systems intrinsically safe by designing complex structures and devising procedures to isolate such equipment from the system or process entirely. However, the uncertainty regarding the effectiveness of such protective measures results in a persisting difficulty in obtaining plant safety certifications and approvals. In this paper, the application of a coded excitation method to make inspection systems intrinsically safe and easily certifiable is explored. Using a pulse compression-based signal processing technique called coded excitation, it has been made possible to achieve non-contact transduction (electromagnetic acoustic transduction and air-coupled transduction) in transmitreceive mode with excitation as low as 0.5 Vpp (peak-to-peak supply voltage). This work reports on the application of coded excitation in bringing down the transduction power requirements for guided ultrasonic wave inspection, thereby making it possible to formulate new inspection applications at very low power, particularly in safety-critical industries.

Optical Material Characterisation of Prepreg CFRP for Improved Composite Inspection

Automated fibre layup techniques are often applied for the production of complex structural components. In order to ensure a sufficient component quality, a subsequent visual inspection is necessary, especially in the aerospace industry. The use of automated optical inspection systems can reduce the inspection effort by up to 50 %. Laser line scan sensors, which capture the topology of the surface, are particularly advantageous for this purpose. These sensors project a laser beam at an angle onto the surface and detect its position via a camera. The optical properties of the observed surface potentially have a great influence on the quality of the recorded data. This is especially relevant for dark or highly scattering materials such as Carbon Fiber Reinforced Plastics (CFRP). For this reason, in this study we investigate the optical reflection and transmission properties of the commonly used Hexel HexPly 8552 IM7 prepreg CFRP in detail. Therefore, we utilise a Gonioreflectometer to investigate such optical characteristics of the material with respect to different fibre orientations, illumination directions and detection angles. In this way, specific scattering information of the material in the hemispherical space are recorded. The major novelty of this research are the findings about the scattering behaviour of the fibre composite material which can be used as a more precise input for the methods of image data quality assessment from our previous research and thus is particularly valuable for developers and users of camera based inspection systems for CFRP components.

An intelligent defect detection system for warp-knitted fabric

In textile factories, the most typical warp-knitted fabric defects include point defects, holes, and color differences. Traditional manual inspection methods are inefficient for detecting these defects. Existing intelligent inspection systems often have a single function. Factories require a real-time inspection system that can detect common defects and color difference. The YOLO (you only look once) neural network is faster than the two-stage neural network and has lower hardware requirements. The system’s color difference detection algorithm compares the color difference between the standard image and the image to be measured and records where the color difference value is exceeded. Finally, the comparison of the factory application proves that the designed system has good real-time performance and accuracy and can meet the fabric inspection requirements of warp-knitted fabric factories.

Analysis of Laser Sensors and Camera Vision in the Shoe Position Inspection System

Inspection systems are currently an evolving field in the industry. The main goal is to provide a picture of the quality of intermediates and products in the production process. The most widespread sensory system is camera equipment. This article describes the implementation of camera devices for checking the location of the upper on the shoe last. The next part of the article deals with the analysis of the application of laser sensors in this task. The results point to the clear advantages of laser sensors in the inspection task of placing the uppers on the shoe’s last. The proposed method defined the resolution of laser scanners according to the type of scanned surface, where the resolution of point cloud ranged from 0.16 to 0.5 mm per point based on equations representing specific points approximated to polynomial regression in specific places, which are defined in this article. Next, two inspection systems were described, where one included further development in the field of automation and industry 4.0 and with a high perspective of development into the future. The main aim of this work is to conduct analyses of sensory systems for inspection systems and their possibilities for further work mainly based on the resolution and quality of obtained data. For instance, dependency on scanning complex surfaces and the achieved resolution of scanned surfaces.

Infer Thermal Information from Visual Information: A Cross Imaging Modality Edge Learning (CIMEL) Framework

The measurement accuracy and reliability of thermography is largely limited by a relatively low spatial-resolution of infrared (IR) cameras in comparison to digital cameras. Using a high-end IR camera to achieve high spatial-resolution can be costly or sometimes infeasible due to the high sample rate required. Therefore, there is a strong demand to improve the quality of IR images, particularly on edges, without upgrading the hardware in the context of surveillance and industrial inspection systems. This paper proposes a novel Conditional Generative Adversarial Networks (CGAN)-based framework to enhance IR edges by learning high-frequency features from corresponding visual images. A dual-discriminator, focusing on edge and content/background, is introduced to guide the cross imaging modality learning procedure of the U-Net generator in high and low frequencies respectively. Results demonstrate that the proposed framework can effectively enhance barely visible edges in IR images without introducing artefacts, meanwhile the content information is well preserved. Different from most similar studies, this method only requires IR images for testing, which will increase the applicability of some scenarios where only one imaging modality is available, such as active thermography.

Utilizing Effective Technology In High-Standard Road Defects Inventory System In Estonia

Road pavement is expected to withstand enormous traffic loads for long time but sooner or later the deterioration reaches levels when its optimal to apply treatment. While easy to measure roughness or rutting in normal traffic speed, defects are in most countries still collected by means of time-consuming visual inspection in low traffic speeds or expensive and difficult- to-use equipment. Also, most visual inspection systems only operate with aggregated inspection data. That makes data-collection expensive and defects-based decision-making inefficient. In Estonia, defects inventory system utilizes high quality panoramic and orthogonal images to enable data collection in traffic speeds and detailed mapping of pavement defects in 10 classes. Defects mapped in full detail means, that location, shape and size of each defect is known and classified data can be effectively used twice in pavement maintenance planning: for section selection planning in road network level when aggregated and for work method selection in design process when analyzed in detail. Combined with measured lidar-based point-cloud data, detailed 3d-basemap saves both road-owner's and road designer’s valuable time in design phase. In period of 2016-2020, around 35000km of state roads were analyzed with one of the most efficient road defects inventory systems in the world. Also, around 25000 km of municipal and forest roads have been captured with same technology covering several pavement types from bicycle paths to multilane streets and motorways. Current presentation discusses outcomes of Estonian defects inventory study in 2020.

Optimization of a CZT photon counting detector for contaminant detection

In the food industry, X-ray inspection systems are utilized to ensure packaged food is free from physical contaminants to maintain a high level of food safety for consumers. However, one of the challenges in the food industry is detecting small, low-density contaminants from packaged food. Cadmium zinc telluride (CZT) photon counting detectors (PCDs) can potentially alleviate this problem given its multi-energy bin capabilities, high spatial resolution and ability to eliminate electronic noise, which is superior to the conventional energy integrating detector (EID). However, the image quality from a CZT PCD can be further improved by applying an optimized energy bin weighting scheme that maximizes energy bin images that provide the largest image contrast and lowest image noise. Therefore, in this work, five contaminant materials embedded in an acrylic phantom were imaged using a CZT PCD while the phantom was in constant motion to mimic food products moving on a conveyor belt. Energy bin optimization was performed by applying an image-based weighting scheme and these results showed contrast-to-noise ratio (CNR) improvements ranging between 1.02–1.91 relative to an equivalent EID acquisition.