Detection of head blight (Fusarium ssp.) in winter wheat by color and multispectral image analyses
Dammer, Karl-Heinz | Möller, Bernd | Rodemann, Bernd | Heppner, Dirk
The use of camera vision to automatically detect head blight (scab) on wheat ears could provide information about the severity of this dangerous disease and help meet future food traceability requirements. Fusarium spp. is dangerous for both human and animal consumption and the ability to monitor symptom location and severity before the harvested product is further processed or stored could help determine whether the grain is fit for human/animal consumption, for bio-conservation, or is completely unusable. To generate various infection levels, field trials were conducted in 2008 and 2009 using wheat varieties with differing levels of susceptibility to the disease; plots were artificially infected with a spore suspension. A color (red, green, and blue) and a multispectral (red, infrared) camera system with real-time image analysis software were developed and compared to detect disease symptoms in the plots. The chlorophyll defect of the infected wheat ears was classified against the image background by setting binarization thresholds. The result was a black and white image. Single pixels or tiny clusters of pixels not belonging to the symptoms were eliminated by setting an area threshold. For both systems, a linear correlation was found between the camera and the visually detected disease levels of the wheat ears in the plots. In the non-infected control plots without disease symptoms, the multispectral system accurately measured “no disease” even though the digital color system detected too much infection (i.e., a false positive). The multispectral system showed a superior calibration capacity. While the color system had to calibrate for each variety, the multispectral system used only one calibration step before starting the measurements.
Показать больше [+] Меньше [-]Ключевые слова АГРОВОК
Библиографическая информация
Эту запись предоставил National Agricultural Library