Determination of pith location along Norway spruce timber boards using one dimensional convolutional neural networks trained on virtual timber boards

Habite, Tadios; Abdeljaber, Osama; Olsson, Anders

doi:10.1016/j.conbuildmat.2022.127129

Cited by 14 publications

(11 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There also appears to be limited literature regarding automated algorithms for locating the pith when only given an image of the end grain of a timber board. Habite, et al (Habite et al, 2020(Habite et al, , 2022 used two neural networks to locate the pith of Norway spruce (Picea abies) timber boards to great success. The first used a conditional generative adversarial network (cGAN) (Habite et al, 2020), and the latter used a one-dimensional convolutional neural network (Habite et al, 2022).…”

Section: Discussionmentioning

confidence: 99%

“…Habite, et al (Habite et al, 2020(Habite et al, , 2022 used two neural networks to locate the pith of Norway spruce (Picea abies) timber boards to great success. The first used a conditional generative adversarial network (cGAN) (Habite et al, 2020), and the latter used a one-dimensional convolutional neural network (Habite et al, 2022). Moreover, it is difficult to estimate the location of the pith using the plastic sheet of concentric circles described by Habite, et al's (Habite et al, 2022(Habite et al, , 2020.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Constructing Virtual Representations of Laminated Timber Products

Grant,

Psaltis,

Shirmohammadi

et al. 2023

Preprint

View full text Add to dashboard Cite

The complex structure of timber has traditionally been difficult to model as it is a highly heterogeneous material. The density (and hence material properties) for structural species such as Pinus radiata (radiata pine) can vary greatly across the growth rings. Numerical simulation methods are becoming more prevalent as a method of predicting moisture migration, stress and strain distributions, and fungal/rot intrusion in timber. These methods require a virtual description of an individual board that captures the growth ring structure present in these timbers. In this work, a low-cost algorithm based on image analysis techniques and spectral segmentation is developed for generating a virtual representation of a cross laminated timber panel. The virtual model is generated as a triangular prismatic mesh where the nodes are aligned on the growth rings, with specific wood material properties allocated to each mesh element The initial step involves identifying the growth ring structure of a single board by analyzing an image of the end grain. The growth rings are then identified using a spectral clustering algorithm. Next, the centre of the tree (pith) is located by using an iterative constrained least-squares algorithm. The density is computed as a proportional length through the growth ring, when provided with an anatomical image of the cellular structure. This density estimate is then assigned to each mesh element resulting in the virtual reconstruction of a singular board. Meshes of multiple timber boards are then combined to produce the final mesh of a laminated timber product.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Constructing Virtual Representations of Laminated Timber Products

Grant,

Psaltis,

Shirmohammadi

et al. 2023

Preprint

View full text Add to dashboard Cite

show abstract

“…In this study such knowledge was obtained manually, on one end of the boards. However, it has recently been shown that location of pith can be determined automatically, in very high speed and with high resolution along the board (Habite et al 2022).…”

Section: Performance Of Models and Indicating Propertiesmentioning

confidence: 99%

“…Habite et al (2020) presented a method to automatically determine location of pith in relation to board cross sections, based on optical scanning of the four longitudinal surfaces of boards and identification of annular ring pattern on images of the surfaces. Faster and more robust methods, based on optical scanning of longitudinal board surfaces in combination with machine learning, were developed by Habite et al (2021) and in particular Habite et al (2022). Thus, it is possible to determine pith location in production speed at sawmills and to utilize this in models of boards.…”

Section: Introductionmentioning

confidence: 99%

Performance of timber board models for prediction of local bending stiffness and strength - with application on Douglas fir sawn timber

Olsson¹,

Pot²,

Viguier³

et al. 2022

WFS

View full text Add to dashboard Cite

Efficient utilization of structural timber requires accurate methods for machine strength grading. One of the most accurate methods presented this far is based on data of local fiber orientation on board surfaces, obtained from laser scanning. In this paper, two potential improvements of this method are examined. The first one consists of replacing a model based on simple integration over cross sections of boards for calculation of local bending stiffness by a 3D solid finite element (FE) model from which local bending stiffness is derived. The second improvement concerns replacement of a simple model for the fiber orientation in the interior of board by a more advanced one taking location of pith and growth direction of knots

show abstract

“…Despite these advancements, the use of machine learning and CNNs in dendrological analyses is limited. Existing works have focused on the automation of pitch detection 40 , 41 and delineating tree ring boundaries and measurements to improve the workflow in dendrochronological studies 42 – 49 . Therefore, this paper aims to explore the performance of CNNs in the resin duct detection task.…”

Section: Introductionmentioning

confidence: 99%

Automatic resin duct detection and measurement from wood core images using convolutional neural networks

Fabijańska

Cahalan

2023

Sci Rep

View full text Add to dashboard Cite

The structure and features of resin ducts provide valuable information about environmental conditions accompanying the growth of trees in the genus Pinus. Therefore analysis of resin duct characteristics has been an increasingly common measurement in dendrochronology. However, the measurement is tedious and time-consuming since it requires thousands of ducts to be manually marked in an image of an enlarged wood surface. Although tools exist to automate some stages of this process, no tool exists to automatically recognize and analyze the resin ducts and standardize them with the tree rings they belong to. This study proposes a new fully automatic pipeline that quantifies the properties of resin ducts in terms of the tree ring area to which they belong. A convolutional neural network underlays the pipeline to detect resin ducts and tree-ring boundaries. Also, a region merging procedure is used to identify connected components corresponding to successive rings. Corresponding ducts and rings are next related to each other. The pipeline was tested on 74 wood images representing five Pinus species. Over 8000 tree-ring boundaries and almost 25,000 resin ducts were analyzed. The proposed method detects resin ducts with a sensitivity of 0.85 and precision of 0.76. The corresponding scores for tree-ring boundary detection are 0.92 and 0.99, respectively.

show abstract

Determination of pith location along Norway spruce timber boards using one dimensional convolutional neural networks trained on virtual timber boards

Cited by 14 publications

References 27 publications

Constructing Virtual Representations of Laminated Timber Products

Constructing Virtual Representations of Laminated Timber Products

Performance of timber board models for prediction of local bending stiffness and strength - with application on Douglas fir sawn timber

Automatic resin duct detection and measurement from wood core images using convolutional neural networks

Contact Info

Product

Resources

About