UruDendro, a public dataset of cross-section images of Pinus taeda

TL;DR

The paper addresses the need for publicly available image datasets to advance automatic tree-ring detection in cross-sections. It introduces UruDendro, a 64-image Pinus taeda dataset with expert ground-truth ring tracings, and presents the Cross-Section Tree-Ring Detection (CS-TRD) algorithm that reconstructs complete rings from full cross-sections using a spider-web of rays, edge chains, and iterative chain merging. CS-TRD achieves a mean F1-score of $89\%$, mean precision $\approx 95\%$, mean recall $\approx 86\%$, and RMSE $\approx 5.27$ px across the dataset, processing images in about $17.3$ seconds on average. The paper also introduces an equivalent-radius metric $r^{eq} = \sqrt{Area/\pi}$ to quantify total ring-area growth, showing strong agreement with manual measurements and enabling holistic diameter-growth assessments. Overall, the dataset and CS-TRD provide a practical, open framework for rapid, low-cost dendrometric analysis and benchmarking of ring-detection methods in conifers.

Abstract

The automatic detection of tree-ring boundaries and other anatomical features using image analysis has progressed substantially over the past decade with advances in machine learning and imagery technology, as well as increasing demands from the dendrochronology community. This paper presents a publicly available database of 64 scanned images of transverse sections of commercially grown Pinus taeda trees from northern Uruguay, ranging from 17 to 24 years old. The collection contains several challenging features for automatic ring detection, including illumination and surface preparation variation, fungal infection (blue stains), knot formation, missing cortex or interruptions in outer rings, and radial cracking. This dataset can be used to develop and test automatic tree ring detection algorithms. This paper presents to the dendrochronology community one such method, Cross-Section Tree-Ring Detection (CS-TRD), which identifies and marks complete annual rings in cross-sections for tree species presenting a clear definition between early and latewood. We compare the CS-TRD performance against the ground truth manual delineation of all rings over the UruDendro dataset. The CS-TRD software identified rings with an average F-score of 89% and RMSE error of 5.27px for the entire database in less than 20 seconds per image. Finally, we propose a robust measure of the ring growth using the \emph{equivalent radius} of a circle having the same area enclosed by the detected tree ring. Overall, this study contributes to the dendrochronologist's toolbox of fast and low-cost methods to automatically detect rings in conifer species, particularly for measuring diameter growth rates and stem transverse area using entire cross-sections.

Figures (11)

• Figure 1: Nine representative examples of the 64 images in the UruDendro data set. Note the variability in color, contrast, the presence of fungus (e.g., image L02b), knots (e.g., images F07b and F03c), and cracks (e.g., images F02e and L03c). The first five images are from the same tree taken at multiple heights.
• Figure 2: Manual tracing of image F08a. (a) image with superimposed marks by three experts; (b) marks, (c) a detail. Marks by experts are violet, red, and yellow. In black is the mean of the three expert marks, and the band in blue is the area span by three standard deviations concerning the mean. Note the variability in the tracing by experts.
• Figure 3: Measuring the error between the expert ring marks and the ground truth for image F07b. a) in green is the ground truth, and in red are the marks produced by the expert. b) areas of influence of the ground truth rings. c) Error, in number of pixels, between the expert marks and the ground truth, see the color bar to interpret error values.
• Figure 4: (a) The whole structure, called spider web, is formed by a center (which corresponds to the cross-section pith), $Nr$rays (in the drawing $Nr=18$) and the rings (concentric curves). In the scheme, the rings are circles, but in practice, they can be (strongly) deformed as long as they don't intersect another ring. Each ray intersects a ring only once in a point called node. The area limited by two consecutive rays and two consecutive rings is named a cell. (b) A curve is a set of connected points (small green dots). Some of those points are the intersection with rays, named nodes (black dots). A chain is a set of connected nodes. In this case, the node$N_i$ is the point$p_n$. (c) Each Chain$Ch_k$ and $Ch_{k+1}$, intersect the rays$R_{m-1}$, $R_{m}$ and $R_{m+1}$ in nodes$N_{i-1}$, $N_{i}$ and $N_{i+1}$. Those rays and chains (as well as the four corresponding nodes) determines cells$C_{l-1}$, $C_{l}$ and $C_{l+1}$.
• Figure 5: Principal steps of the CS-TRD tree-ring detection algorithm: (a) original image, (b) background extraction, (c) pre-processed image (resized, equalized, and converted to a grayscale image), (d) the output of the Canny Devernay edge detector, (e) edges filtered by the direction of the gradient, (f) set of detected chains, (g) connected chains, (h) post-processed chains and (i) detected tree-rings.