Scribd
Upload
64 views3 pages

Identification of Seeds

The document describes rule sets for automated identification of seeds and seedlings from aerial images using eCognition software. For seed identification, it involves segmentation, brightness thresholding, shape analysis to identify seeds. For target seedling identification of L. angustifolia, it uses segmentation, spectral difference analysis, texture and shape parameters to identify definite, maybe and non-target classes. Figures S1 and S2 show example outputs highlighting the identified classes and missed targets.

Uploaded by

Lcelectronic Maracaibo
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
64 views3 pages

Identification of Seeds

The document describes rule sets for automated identification of seeds and seedlings from aerial images using eCognition software. For seed identification, it involves segmentation, brightness thresholding, shape analysis to identify seeds. For target seedling identification of L. angustifolia, it uses segmentation, spectral difference analysis, texture and shape parameters to identify definite, maybe and non-target classes. Figures S1 and S2 show example outputs highlighting the identified classes and missed targets.

Uploaded by

Lcelectronic Maracaibo
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Supplementary S1.

Full eCognition rule sets for automated seed and seedling identification
from Buters et al. 2019.

Identification of seeds:
1: Multiresolution segmentation (scale parameter 10, shape: 0.4, compact: 0.7) creating ‘New
Level’
2: Assign class – unclassified with Total diff to scene ≥100 at New Level: Seed
3: Merge – Seed with Brightness ≥0 at New Level: merge region
4: Assign class – Seed with Length/Width >1.5 at New Level: unclassified
5: Assign class – Seed with Area <0.8 cm2 at New Level: unclassified
6: Assign class – Seed with Roundness ≥0.8 at New Level: unclassified

Identification of target seedlings (L. angustifolia):


1: Multiresolution segmentation (scale parameter: 20, shape: 0.3, compact: 0.7) creating ‘New
level’
2: Assign class – unclassified with green ratio* >0.36 at New Level: target
3: Spectral difference segmentation – target at New level: spectral difference 30
4: Assign class – target with Mean diff. to neighbours DEM (0) ≥0.007 at New Level: definitely
target
5: Assign class – target with HSI Transformation Saturation (R = Layer 3, G = Layer 2, B = Layer
1) <0.1 at New level: unclassified
6: Assign class – target with HSI Transformation Hue (R = Layer 1, G = Layer 2, B = Layer 3)
≥0.2 and HSI Transformation Saturation (R = Layer 1, G = Layer 2, B = Layer 3) ≥0.2 at New
level: maybe target
7: Assign class – maybe target, target with Perimeter/Width >18 at New level: unclassified
8: Assign class – maybe target with Compactness <2.5 at New level: definitely target
9: Assign class – maybe target, target with TGI <21 at New level: unclassified
10: Assign class – maybe target, target with TGI ≥26 at New level: definitely target
11: Pixel based object resizing – maybe target at New level: shrink using maybe target where rel.
area of object pixels in 5 x 5 ≥0.2
12: Assign class – definitely target, maybe target, target with Length/Width ≥5.5 at New level:
unclassified
13: Assign class – definitely target, maybe target, target with Area ≤16 cm2 at New level:
unclassified
14: Assign class – target with Brightness ≥0 at New level: unclassified
15: Assign class – maybe target with Brightness ≥0 at New level: unclassified

*Green ratio= Mean Green/(Mean Red + Mean Green + Mean Blue)

Identification of non-target grass cover (applied following Identification of target seedlings):


1: Assign class – unclassified with TGI >6.5 and green ratio >0.35 at New level: non-target grass
Figure S1. Example output image of a processed target area identifying seeds from eCognition
rulesets, highlighting classes ‘seeds’ (red), as well as ‘missed target’ (blue) manually identified
post-processing (A), with corresponding unprocessed image target area (B) in which target seeds
have been manually identified (annotated arrows). Image taken from an altitude of 5 m with a DJI
Phantom 4 Pro UAV.
Figure S2. Example output image of a processed target area identifying seedlings from eCognition
rulesets, highlighting classes ‘definitely target’ (yellow) and ‘non-target grasses’ (red), as well as
‘missed target’ (blue) manually identified post-processing (A), with corresponding unprocessed
image target area (B) in which target seedlings have been manually identified (annotated arrows).
Image taken from an altitude of 5 m with a DJI Phantom 4 Pro UAV.

You might also like