Spectral Imaging, or imaging spectroscopy is a widespread sensor technology used in precision agriculture, horticulture and plant phenotyping. From cameras providing just a few spectral bands on drones, to cameras with a large number of bands, often referred to as hyperspectral cameras on field vehicles or in greenhouses. This technology enables plant scientists to quantify the composition of agricultural products, and detection of plant stresses and diseases in an early stage.
Traditionally, analysis of spectral image data is performed using classical machine learning on the spectral or image components. Nowadays convolutional neural or deep learning networks are becoming immensely popular particularly for RGB image data using three input layers - a large number of pretrained networks are available. For spectral image data, with a large number of input bands these networks do not work out of the box and need to be adapted. Also, the adaption of pre-trained weights based on RGB data needs attention. The goal of this Research Topic is to present a platform for researchers where solutions for the application of existing deep learning networks for spectral image data, and the new development of deep learning networks for spectral imaging, can be critically assessed and shared. This is essential for more effective detection and classification methods for precision agriculture, horticulture and plant phenotyping can be further developed.
This Research Topic focuses on new developments in convolutional neural networks and deep learning for spectral image data, particularly in agriculture sciences. The topic welcomes, but is not limited to, Original Research and Reviews related to the deep learning for spectral imaging in precision agriculture, horticulture and plant phenotyping areas:
• Plant disease detection
• Measuring biotic and abiotic plant stresses
• Postharvest quality assessment
• Integrated pest management
• Drone and field-based plant phenotyping
• Seed identification and classification
Spectral Imaging, or imaging spectroscopy is a widespread sensor technology used in precision agriculture, horticulture and plant phenotyping. From cameras providing just a few spectral bands on drones, to cameras with a large number of bands, often referred to as hyperspectral cameras on field vehicles or in greenhouses. This technology enables plant scientists to quantify the composition of agricultural products, and detection of plant stresses and diseases in an early stage.
Traditionally, analysis of spectral image data is performed using classical machine learning on the spectral or image components. Nowadays convolutional neural or deep learning networks are becoming immensely popular particularly for RGB image data using three input layers - a large number of pretrained networks are available. For spectral image data, with a large number of input bands these networks do not work out of the box and need to be adapted. Also, the adaption of pre-trained weights based on RGB data needs attention. The goal of this Research Topic is to present a platform for researchers where solutions for the application of existing deep learning networks for spectral image data, and the new development of deep learning networks for spectral imaging, can be critically assessed and shared. This is essential for more effective detection and classification methods for precision agriculture, horticulture and plant phenotyping can be further developed.
This Research Topic focuses on new developments in convolutional neural networks and deep learning for spectral image data, particularly in agriculture sciences. The topic welcomes, but is not limited to, Original Research and Reviews related to the deep learning for spectral imaging in precision agriculture, horticulture and plant phenotyping areas:
• Plant disease detection
• Measuring biotic and abiotic plant stresses
• Postharvest quality assessment
• Integrated pest management
• Drone and field-based plant phenotyping
• Seed identification and classification