Automated Defect Detection in Leather Products

Leather remains a cornerstone of global manufacturing, spanning footwear, furniture, automotive upholstery and luxury goods. Ensuring consistent surface quality is critical both for economic value and consumer confidence, yet traditional manual inspection is labour‐intensive, subjective and prone to inconsistency. Automated defect detection systems integrate high‐resolution imaging—ranging from visible‐light cameras to hyperspectral sensors—with advanced algorithms to identify, localise and classify anomalies such as scratches, insect bites, wrinkles and colour variations. A typical workflow comprises image acquisition under controlled illumination, preprocessing for noise reduction and contrast enhancement, followed by feature extraction and decision making. Classical image‐analysis techniques (edge detectors, texture descriptors) have gradually given way to machine‐learning and deep‐learning approaches, in which convolutional neural networks (CNNs) and, more recently, vision transformers (ViTs) learn discriminative representations directly from data. These systems deliver real‐time grading, reduce material waste, standardise quality control across production lines and support Industry 4.0 objectives of flexible, data‐driven manufacturing. Key challenges include the heterogeneous nature of leather grain, varying defect scales, limited annotated datasets and the need for computationally efficient models deployable at the production edge.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Automated Defect Detection in Leather Products publication trend

The graph below shows the total number of articles in automated defect detection in leather products across all publications each year (not limited to Nature Index journals).

Technical terms

Machine vision: Automated acquisition and analysis of images to extract information about physical objects or scenes.

Convolutional neural network (CNN): A class of deep learning models that applies convolutional filters to learn spatial hierarchies of features from images.

Vision transformer (ViT): An attention-based architecture that divides an image into patches and models relationships between them using self-attention layers.

Hyperspectral imaging (HSI): A technique capturing image data at many narrow spectral bands to reveal chemical or material differences not visible in standard colour images.

Attention mechanism: A model component that weights input features according to relevance, enhancing salient patterns for tasks such as detection or classification.

Anomaly detection: The process of identifying patterns in data that deviate from the expected “normal” behaviour, often using one-class or unsupervised methods.

References

1. A Systematic Review of Machine-Vision-Based Leather Surface Defect Inspection. Electronics (2022).
2. Surface Defect Detection of Wet-Blue Leather Using Hyperspectral Imaging. IEEE Access (2021).
3. Vision Transformers for Anomaly Detection and Localisation in Leather Surface Defect Classification Based on Low-Resolution Images and a Small Dataset. Applied Sciences (2023).
4. Leather Defect Detection Based on Improved YOLOv8 Model. Applied Sciences (2024).
5. Deep learning and machine learning neural network approaches for multi class leather texture defect classification and segmentation. Collagen and Leather (2022).