Triangle Scan Method (TSM)

The triangle scan method is a 2-step process which determines the central crop row from a semantic skeleton segmentation of a crop row. The first step (Anchor Scan) and the second step (Line Scan) scan for the topmost (Anchor Point) and the lowermost ($P_{r}$) points of the central crop row respectively. The anchor point $A$ and point $P_{r}$ always lie on the top and bottom edges of the image respectively. Please refer the TSM documentation for further details.

Figure 1: Regions of interest for anchor scan and line scan. Anchor Scan ROI: RED, Line Scan ROI: Green.

Usage

Prerequisits

1. Install dependencies: [OpenCV, NumPy, SK-Image, Seaborn, Pandas, Matplotlib, Glob, Argparse]

pip install opencv-python numpy scikit-image seaborn pandas matplotlib glob3 argparse

2. Clone the repository.

git clone https://github.com/Rajitha159/TSM.git

3. Copy the RGB image files to "rgb" folder and predicted crop row masks to "mask" folder(all images must be resized to $512 \times 512$).
4. Run the following command to generate crop row masks and save them to "out" folder.

python triangle_scan_rgb.py

5. [Optional] Run the code with parameters in needed. Parameter descriptions are in table below. Example:

python triangle_scan_rgb.py --file_type=".png"

Parameters

Parameter	Description
--file_type	Specify the file type for image files. Default is .jpg.
--A	Specify the standard anchor point. Default is 277.
--B	Specify the begin point (B) for line scans. Default is 200.
--C	Specify the cease point (C) for line scans. Default is 450.
--Amin	Specify the anchor scans starting point. Default is 100.
--Amax	Specify the anchor scans ending point. Default is 350.
--s	Specify the anchor scans ROI (Region of Interest) height. Default is 0.2.
--scan_period	Specify the scan period for anchor and line scans. Default is 1.
--filter_enable	Enable the complementary filters for scanning in continuous sequential images. Default is False.
--anchor_filter	Specify the complementary filter strength for anchor scans. Default is 0.95.
--line_filter	Specify the complementary filter strength for line scans. Default is 0.95.

Citation

@article{de2024vision,
  title={Vision based crop row navigation under varying field conditions in arable fields},
  author={de Silva, Rajitha and Cielniak, Grzegorz and Gao, Junfeng},
  journal={Computers and Electronics in Agriculture},
  volume={217},
  pages={108581},
  year={2024},
  publisher={Elsevier}
}

Link to full paper: Vision based Crop Row Navigation under Varying Field Conditions in Arable Fields

Licence: Creative Commons Attribution-NonCommercial-NoDerivatives 4.0