A Functional Survey of AI-Based Vision Systems for Industrial Applications: Safety, Quality, and Productivity

Computer Vision and Image Understanding

Dependable Person Detection using AI in Industrial Environments

Suggestions for ensuring safe person detection using AI in industrial environments are offered, including suggestions for ensuring safe person detection using AI in industrial environments.

Artificial Intelligence and Computer Vision for Automated Aircraft Defect Detection

The inspection of aeronautical components requires compliance with extremely high safety and quality standards, where early detection of defects and verification of surface conditions are critical for aircraft reliability. The use of robots equipped with computer vision enables the automation of these inspections, improving coverage, consistency, and safety compared to manual methods, while enhancing overall operational efficiency. The incorporation of artificial intelligence further enhances these systems by enabling automatic defect detection, classification according to severity, and real-time inspections, reducing human errors and streamlining the inspection process. This work presents a technically viable solution not only for defect detection but also for evaluating the cleanliness or contamination of aeronautical surfaces, using non-destructive methods based on computer vision, profilometry, and artificial intelligence, grounded in image acquisition and analysis. Validation at demonstrator scale confirmed the system’s ability to meet the initially defined specifications, demonstrating reliable detection of both defects and organic residues, including those caused by impacts and insect accumulation on aircraft surfaces. This approach combines automation, precision, and traceability, representing a significant step forward toward advanced and safe inspection practices in the aerospace industry.

Restructuring Industrial Cognition: An Ai-Enabled Perceptual Alignment Framework For Quality Inspection

Artificial intelligence (AI) is increasingly embedded in manufacturing systems, reshaping not only operational efficiency but also the epistemological foundations of quality decision-making (Brynjolfsson & McAfee, 2017; Porter & Heppelmann, 2014). While prior studies have emphasized the technical advantages of AI-driven inspection systems, limited attention has been given to how such systems transform perceptual judgment and redefine the boundary between human and machine cognition in industrial contexts (Gambino et al., 2020; Longoni et al., 2019). Addressing this gap, this study proposes an AI-enabled perceptual alignment framework for spherical surface inspection, integrating skeleton-based feature matching with three-dimensional (3D) pose estimation to mitigate perceptual uncertainty in high-precision manufacturing environments. Drawing on impression integration theory (Anderson, 1981; Kunda & Thagard, 1996), this research conceptualizes quality inspection as a perceptual integration process in which cognitive and visual cues are synthesized to form defect judgments. Traditional inspection systems—heavily reliant on human operators—are prone to variability due to fatigue, subjective bias and the inherent complexity of curved surfaces. Using a real-world case of golf ball surface inspection, this study demonstrates how AI mediated perception can standardize judgment by reconstructing object orientation, correcting geometric distortions and enabling consistent feature recognition under variable conditions. Empirical findings indicate that the proposed framework significantly reduces inspection variability, lowers false detection rates and enhances yield stability. More importantly, the results reveal a structural shift from human-centric to AI-mediated decision-making, in which perceptual authority is redistributed from individual operators to algorithmic systems. This transformation contributes to the emergence of more data-driven forms of quality governance and more traceable decision processes within manufacturing systems. Theoretically, this study extends impression integration theory to human–AI interaction in industrial settings by demonstrating how AI systems not only replicate but also recalibrate perceptual judgment processes. Practically, it offers a scalable and modular solution for spherical object inspection with broader applicability across precision manufacturing domains. By positioning AI as a mediator of perception, this research provides new insights into the evolving role of intelligent systems in shaping industrial cognition and decision-making.

Artificial Intelligence in Industrial Production: A Survey of concepts, technologies & Practical Application in an Intelligent Production Environment

The increasing digitalization of industrial production systems in the context of Industry 4.0 is leading to a growing use of data-driven and intelligent technologies in manufacturing and assembly environments. In particular, Artificial Intelligence (AI), machine learning, and deep learning methods based on neural networks open up new possibilities for the automation of complex decision-making and inspection processes. One of the central application areas in this context is AI-based image processing for visual inspection and quality control. This paper provides a structured and comprehensive overview of the fundamental concepts, technologies, and methods of Artificial Intelligence in the context of industrial production. Among other aspects, relevant neural network architectures with a focus on industrial image processing, typical training and optimization procedures, data preprocessing, and key challenges are addressed. In addition, common application areas of AI in Smart Factory environments are systematically presented. Finally, the paper presents current research at Hochschule Bochum, in which the practical implementation of these technologies using developed neural networks for automated visual quality control in a gearbox assembly process is demonstrated. Lastly, a comparative approach between the AI-based image processing solution and a conventional rule-based machine vision system is presented.

Examining the Visual Capabilities of Multimodal Large Language Models for Automotive Applications

Automatic recognition and classification of vehicle damages is an important research direction in modern computer vision and artificial intelligence, playing an increasingly significant role in industrial and practical applications. Traditional computer vision-based approaches can recognize and classify objects with high accuracy; however, achieving task-specific performance typically requires large amounts of annotated data, time-consuming training or fine-tuning, and extensive parameter optimization. This process is not only resource- and cost-intensive but also limits the rapid adaptability of the technology. The aim of this research is to investigate how effectively the latest Multimodal Large Language Models (MLLMs) can recognize types of vehicle damage in a zero-shot setting, i.e., without fine-tuning, and to evaluate how their performance can be further improved through prompt engineering and fewshot prompting. MLLMs have the advantage of being able to provide multiple forms of information from a single query and supplement their outputs with natural-language explanations. In contrast, traditional models are generally designed to perform only one predefined task. Therefore, within the framework of this project, the performance of a fine-tuned YOLO-based computer vision model is compared with that of MLLMs in vehicle damage classification. This comparison highlights a modern, data-efficient approach that achieves competitive performance through prompt engineering and in-context learning, eliminating the need for additional model training and opening new directions for automotive applications.