cereal and broadleaf crops. The proximity based hyperspectral camera yielded hyperspectral resolution with high spatial resolution, enabling considerable spectral and spatial separation between crop and weed. In a different approach, Eddy et al. (2013) have investigated the feasibility of using reduced hyperspectral bandsets and ANN classification for discriminating between crop-field pea (Pisum sativum), spring wheat (Triticum aestivum), canola (Brassica napus) and weed-wild oat (Avena fatua), redroot pigweed (Amaranthus retroflexus). Reduced sets of narrow wave bands were created using Principal Component Analysis and Stepwise Discriminant Analysis with experimental results showing that plant discrimination using an ANN classifier was feasible and could provide considerable computational savings due to the reduced data dimensionality. The drawback however, was the high overhead required to train the classifier for successful operation.
This paper reports on recent results obtained from research into the development of an advanced proof-of-concept real-time plant discrimination system based on discrete spectral reflectance measurements for green-from-green plant discrimination. The developed system is tested for the discrimination of Anthurium (Anthurium andraeanum) from Sunkisses (Ipomoea batatas var. sunkisses) and Dandelion (Taraxacum officinale).
Experimental results show that practical green-from-green discrimination at a farming vehicle speed of 3 km/h can be achieved. At higher speeds, due to identified hardware limitations, the discrimination capability and accuracy declines.
cereal and broadleaf crops. The proximity based hyperspectral camera yielded hyperspectral resolution with high spatial resolution, enabling considerable spectral and spatial separation between crop and weed. In a different approach, Eddy et al. (2013) have investigated the feasibility of using reduced hyperspectral bandsets and ANN classification for discriminating between crop-field pea (Pisum sativum), spring wheat (Triticum aestivum), canola (Brassica napus) and weed-wild oat (Avena fatua), redroot pigweed (Amaranthus retroflexus). Reduced sets of narrow wave bands were created using Principal Component Analysis and Stepwise Discriminant Analysis with experimental results showing that plant discrimination using an ANN classifier was feasible and could provide considerable computational savings due to the reduced data dimensionality. The drawback however, was the high overhead required to train the classifier for successful operation.This paper reports on recent results obtained from research into the development of an advanced proof-of-concept real-time plant discrimination system based on discrete spectral reflectance measurements for green-from-green plant discrimination. The developed system is tested for the discrimination of Anthurium (Anthurium andraeanum) from Sunkisses (Ipomoea batatas var. sunkisses) and Dandelion (Taraxacum officinale).Experimental results show that practical green-from-green discrimination at a farming vehicle speed of 3 km/h can be achieved. At higher speeds, due to identified hardware limitations, the discrimination capability and accuracy declines.
การแปล กรุณารอสักครู่..