This is the graduation project for the Master of Science in Forestry in UNB. It uses the spherical camera (Rioch Theta S) to estimate forest attributes under canopy. This project has three parts:

1. Estimating Stand Basal Area from spherical images.
2. Estimating Canopy Structural Fractions (plant, sky, foliage, etc.) from spherical images
3. Estimating Individual tree Attributes (DBH, Height, spatial location) from two spherical image pairs.

See this project: https://github.com/HowcanoeWang/Panorama2BasalArea

The workflow of two approaches to estimate plant fraction (PF) from spherical photos. The left blue workflow is the hemispherical approach (HPF) which converts original cylindrical images to hemispherical images first then apply algorithms commonly used in hemispherical images. The right green one is the cylindrical approach (CPF) which directly calculate the PF value on original cylindrical images without image converting.

See Plant Fraction Folders，the item goes:

Plant Fraction/
|- config.py   # The HSV theshold and path settings
|- Hemispherical/   # Hemispheical Approach in Thesis Chapter 3
|   |- Conversed_57.5/   # The output folder for generated fisheye images.
|	|- converse.py   # Convert Raw image to Fisheye image with distortion calibrated
|	|- classify_all_fisheye.py   # The model for HSV classification
|   |- plant_fraction_hemi.py   # Operate HSV classification for all fisheye images.
|- Cylindrical/
	|- plant_fraction_cyli.py   # The model for HSV classification for raw image
    |- classify_all_cylindircal   # Operate HSV classification for all images

Currently, no executable app has been packed.

The GUI to mark key points (base and top of tree) in spherical image pairs. However, limited by schedule, the database hasn't been developed, all the calculating data needs to be pasted to the Excel file (DataTemplate.xlsx)

Operation Steps:

1. Run app.py scripts or app.exe downloaded in this link.
2. Load spherical images (OpenImg button) at 1.6m and 2.6m for left panel and right panel respectively.
3. Mark ground control points (click once on each image, e.g. plot center)
4. Press (Convert) in 1.6m img, and paste result to the first column in Plot Sheet
5. Press (Convert) in 2.6m img, and paste result to the next column in Plot Sheet
6. Press N to start marking a new tree.
7. Following this order to mark key points (Using windows magnifier tools if it is hard to see clearly):
   1. 1.6m img tree base
   2. 2.6m img tree The model for HSV classification
   3. (the horizonal red line will be locked at 1.3m)
   4. left point of DBH at 1.6m img
   5. right point of DBH at 1.6m img
   6. left point of DBH at 2.6m img
   7. right point of DBH at 2.6m img
   8. 1.6m img tree top
   9. 2.6m img tree base
   10. Paste result to a row of Tree Sheet
   11. repeat previous steps for all trees in this spherical image pair
8. Reload next plot images.

Future work: Integrate all the previous functions into one app.

icon url: