176
produced the final signature for each sample. The classification accuracy was calculated using the ERDAS
®
Imagine Accuracy Assessment feature and 200 randomly
distributed points across the chosen 7-band image for both unsupervised and supervised classifications.
We conducted a multivariate analysis between full spectral signatures of samples
to determine whether they were significantly different using all 7 bands (spectral
separability). Non-target species’ signatures were compared to R. ponticum, using
multivariate analysis, to determine which bands showed the significant variation
within samples.
Results
Distortion Effects and Removal
Non-gimbal mounted cameras suffered from vibration wave-distortions and blur on
the image outputs, meaning that many pictures were unsuitable for mosaicking.
Identifying errors in the images and removing the distortion was time consuming:
gimbal mounting and the removal of fisheye lenses would fully resolve these issues.
Spectral Analysis
R. ponticum was shown in inter-species pairwise comparisons to show the greatest
spectral separability at 540 (p < 0.05) and least at 550 (p > 0.05) using multivariate
analysis. Intra-species difference between two populations growing in different
locations was also significant (p = 0.01) in the same pairwise comparison. Ivy was
not significantly different to Larch and R. ponticum in all but the 540 and 750
Fig. 5 Checkerboard created in virtual workspace for spectral signature analysis and
classification
A. Sanders
produced the final signature for each sample. The classification accuracy was calculated using the ERDAS
®
Imagine Accuracy Assessment feature and 200 randomly
distributed points across the chosen 7-band image for both unsupervised and supervised classifications.
We conducted a multivariate analysis between full spectral signatures of samples
to determine whether they were significantly different using all 7 bands (spectral
separability). Non-target species’ signatures were compared to R. ponticum, using
multivariate analysis, to determine which bands showed the significant variation
within samples.
Results
Distortion Effects and Removal
Non-gimbal mounted cameras suffered from vibration wave-distortions and blur on
the image outputs, meaning that many pictures were unsuitable for mosaicking.
Identifying errors in the images and removing the distortion was time consuming:
gimbal mounting and the removal of fisheye lenses would fully resolve these issues.
Spectral Analysis
R. ponticum was shown in inter-species pairwise comparisons to show the greatest
spectral separability at 540 (p < 0.05) and least at 550 (p > 0.05) using multivariate
analysis. Intra-species difference between two populations growing in different
locations was also significant (p = 0.01) in the same pairwise comparison. Ivy was
not significantly different to Larch and R. ponticum in all but the 540 and 750
Fig. 5 Checkerboard created in virtual workspace for spectral signature analysis and
classification
A. Sanders