Received 4 April 2025, accepted 17 April 2025, date of publication 21 April 2025, date of current version 5 May 2025. Digital Object Identifier 10.1109/ACCESS.2025.3562873
Integration of Convolutional Neural Network and Image Processing for Pulp Fibril Detection and Measurement TANACHOT CHIRAKITSAKUL 1,2 , PAKAKET WATTUYA 1,2 , PHICHIT SOMBOON 3 , PANTHIRA JANSAKRA 3 , AND CHAKRIT WATCHAROPAS 1,2 , (Member, IEEE) 1 Department of Computer Science, Kasetsart University, Bangkok 10900, Thailand 2 Artificial Intelligence Technology and Innovation Center for Health, Faculty of Science, Kasetsart University, Bangkok 10900, Thailand 3 Department of Forest Products, Kasetsart University, Bangkok 10900, Thailand Corresponding author: Pakaket Wattuya (pakaket.w@ku.th) ABSTRACT The fibrillation index is a critical metric in paper manufacturing, quantifying the degree of fibrillation achieved during the pulp refining process. Optimizing this metric enhances both paper quality and production efficiency. However, traditional measurement methods—such as manual visual examination of pulp samples under microscopy—are slow, error-prone, and labor-intensive, limiting their scalability in industrial applications. This study proposes a novel method that integrates deep learning with image processing techniques to automate fibril detection and fibrillation index computation. The proposed method leverages the discriminative capabilities of convolutional neural networks (CNNs) with adaptive image processing techniques to overcome key challenges such as low contrast, image noise, and variability in fibril morphology. The patch-based classification approach effectively filters out irrelevant objects, especially those whose features visually resemble fibrils, thus improving fibril segmentation accuracy. The method was comprehensively validated against expert-labeled ground truth images and achieved a promising average error rate of 0.4494 ± 0.4187. Experimental results also demonstrate the strong robustness of the proposed method, with consistent performance across diverse refining conditions and image qualities, making it suitable for real-world application in the pulp and paper industry. Furthermore, this study paves the way for broader applications in materials science and biomedical imaging, where precise feature detection in microscopic images is essential.
INDEX TERMS Convolutional neural network, deep learning, image segmentation, fibril detection, pulp fibril analysis, fibrillation index.
I. INTRODUCTION The global pulp and paper industry has witnessed significant growth due to the rising demand for sustainable packaging and hygiene products, even as traditional paper products like newspapers continue to decline. At the core of enhancing product quality in this sector lies the pulp refining process, during which fibers undergo mechanical treatment to improve their bonding potential and product quality. The major pulp fiber development from this process is the external
fibrillation, where the outer layers of pulp fibers are stripped into thread-like structures called fibrils (see Figure 1). Fibrils significantly enhance paper strength, surface properties, and bonding capabilities, making their analysis indispensable for optimizing paper production and refining operations. The fibrillation index, a key metric in pulp and paper manufacturing, quantifies the degree of fibrillation by calculating the ratio of fibril area to fiber area in microscopic images [1]. This metric is essential for understanding refining efficiency and tailoring the mechanical properties of paper products [2], [3], [4], [5], [6], [7], [8]. This metric also plays a vital role in sustainability by improving the efficiency of
The associate editor coordinating the review of this manuscript and approving it for publication was Prakasam Periasamy .
2025 The Authors. This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/
74634
VOLUME 13, 2025
Made with FlippingBook. PDF to flipbook with ease