CS-TRD: a Cross Sections Tree Ring Detection method

Accepted at Image Processing On Line (IPOL) 2025.

Repository for the paper "CS-TRD: a Cross Sections Tree Ring Detection method".

Demo: DEMO.

UruDendro ImageSet: UruDendro. ArXiv paper: ArXiv.

---

---

---

Version 1.0 Last update: 23/05/2025

Authors:

• Henry Marichal, [email protected]
• Diego Passarella, [email protected]
• Gregory Randall, [email protected]

Get started

1. Folders

All the python source files are in lib/ folder. Dependencies are in externas/ folder. Default configuration files are in config/ folder.

Paper

Algorithm 1 is implemented in the file main.py

Line 1 of Algorithm 1 is implemented in file lib/preprocessing.py

Canny-Devernay implementation [10] is in externas/devernay_1.0. Python wrapper is in lib/canny_devernay_edge_detector.py

Line 3 of Algorithm 1 is implemented in the file lib/filter_edges.py

Algorithm 2 is implemented in the file lib/sampling.py

Algorithm 3, 4, 5 are implemented in the file lib/merge_chains.py

Algorithm 6 is implemented in the file lib/basic_properties.py

Algorithm 7 is implemented in the file lib/postprocessing.py

Installation

Setup environment

Conda

conda create --name ipol python==3.11
conda activate ipol
conda install -n ipol -c conda-forge geos
conda install -n ipol -c anaconda cmake 
conda install -n ipol pip

APT

sudo apt-get update && sudo apt-get install -y libgeos-dev=3.8.0-1build1 cmake=3.16.3-1ubuntu1.20.04.1

Install the package

python setup.py install
pip install -r requirements.txt

Examples of usage

Import the module

from cross_section_tree_ring_detection.cross_section_tree_ring_detection import TreeRingDetection
from cross_section_tree_ring_detection.io import load_image

args =  dict(cy=1264, cx=1204, sigma=3, th_low=5, th_high=20,
        height=1500, width=1500, alpha=30, nr=360,
        mc=2)

im_in = load_image('input/F02c.png')
res = TreeRingDetection(im_in, **args)

rings_point = res[-1]

CLI

python main.py --input input/F02c.png --cy 1264 --cx 1204  --output_dir ./output --root ./

If you want to run the algorithm generating intermediate results you can use the flag --save_imgs

python main.py --input input/F02c.png --cy 1264 --cx 1204  --output_dir ./output --root ./ --save_imgs 1

Automatic center detection

Detecting pith center automatically can be done using software from IPOL paper "Ant Colony Optimization for Estimating Pith Position on Images of Tree Log Ends" IPOL.

Automatic background removal

Install repository uruDendro. Then,

from urudendro.remove_salient_object import remove_salient_object
IMAGE_PATH = 'input/F02c.jpg'
OUTPUT_PATH = 'input/F02c_segmented.jpg'
remove_salient_object(IMAGE_PATH, OUTPUT_PATH)

Metric Computation

Install repository uruDendro. Then, Tree ring Detection evaluation metrics used by UruDendro. The evaluation code provided here can be used to obtain results on the publicly available UruDendro dataset. It computes multiple metrics described below.

from urudendro.metric_influence_area import main as metric
    
DETECTION_FILENAME = 'path/to/detection/file'
GROUND_TRUTH_FILENAME = 'path/to/ground/truth/file'
IMAGE_FILENAME = 'path/to/image/file'
CX = 0
CY = 0
THRESHOLD = 0.5
OUTPUT_DIR = 'path/to/output/directory'

P, R, F, RMSE, TP, FP, TN, FN = metric(DETECTION_FILENAME, GROUND_TRUTH_FILENAME, IMAGE_FILENAME, CX, CY, THRESHOLD, OUTPUT_DIR)

where DETECTION_FILENAME is the path to the detection file, GROUND_TRUTH_FILENAME is the path to the ground truth file, IMAGE_FILENAME is the path to the image file, CX is the x pith coordinate in pixels, CY is the y pith coordinate in pixel, THRESHOLD is the threshold to consider a detection as valid (between 0 and 1) and OUTPUT_DIR is the path to the results output directory.

Docker Container

You can run the algorithm in a docker container.

Pull the image

docker pull hmarichal/cstrd:v1.0

Run the container

In order to run the container you need to mount a volume with the data you want to process (YOUR_DATA_FOLDER). Results will be stored in the mounted volume. Run the following command:

docker run -v YOUR_DATA_FOLDER:/workdir/bin/output -it hmarichal/cstrd:v1.0 / 
 python main.py --input YOUR_DATA_FOLDER/image_path --cy 1264 --cx 1204 /
 --output_dir ./output --root ./ --save_imgs 1

Build the image

 docker build -f .ipol/Dockerfile . -t hmarichal/cstrd:v1.0

Citation

@misc{marichal2023cstrd,
      title={CS-TRD: a Cross Sections Tree Ring Detection method}, 
      author={Henry Marichal and Diego Passarella and Gregory Randall},
      year={2023},
      eprint={2305.10809},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

License

License for th source code: MIT