Spectroscopy and imaging methods have been successively applied to crop stress determination due to the advantages of rapid speed, non-invasive approach, and low measurement cost, and the introduction of machine learning methods provides reliable support for intelligent and automatic analysis.
However, with the in-depth study of crop stress, the combination of spectroscopy, imaging and machine learning has encountered some challenges. Firstly, single spectroscopy or imaging techniques may be inadequate in stress analysis. For example, RGB only percepts the external characteristics, while infrared spectroscopy senses the internal composition changes in shallow layers of crops. Fusion or improvement of spectroscopy and imaging techniques can alleviate the problems. Secondly, the spectroscopic and image features of various crop stresses are different, and there is a strong coupling relationship between stress and crop growth, such as growth stage and phenotype, rainfall, temperature and humidity of growth environment, etc. Therefore, it is necessary to establish a systematic stress diagnosis mechanism. Meanwhile, spectra and images are of various sources and heterogeneous data attributes, but the existing separate and independent processing approaches are difficult to obtain good performance, and the machine learning methods with simple architecture generally perform poorly in extracting high-quality features from the data with heterogeneous attributes. Additionally, for the large-scale analysis of crop stress, the instruments of spectroscopy and imaging need to be loaded on unmanned aerial vehicles and mobile vehicles, and the installation, measurement method, and data transmission and processing method should be explored.
This Research Topic aims to develop novel spectroscopy-image methods and machine learning to solve the crop stress detection problems. We welcome submissions of original research articles, reviews, and methods, including (but not limited to) research on the following sub-themes:
• Spectral and image databases for crop stress
• Fusion of spectroscopy and imaging for stress analysis
• New spectroscopy and imaging techniques
• Improvement of machine learning and deep learning
• High throughput sensing system of crop stress
• Wide-scale and mobile perception of crop stress
Spectroscopy and imaging methods have been successively applied to crop stress determination due to the advantages of rapid speed, non-invasive approach, and low measurement cost, and the introduction of machine learning methods provides reliable support for intelligent and automatic analysis.
However, with the in-depth study of crop stress, the combination of spectroscopy, imaging and machine learning has encountered some challenges. Firstly, single spectroscopy or imaging techniques may be inadequate in stress analysis. For example, RGB only percepts the external characteristics, while infrared spectroscopy senses the internal composition changes in shallow layers of crops. Fusion or improvement of spectroscopy and imaging techniques can alleviate the problems. Secondly, the spectroscopic and image features of various crop stresses are different, and there is a strong coupling relationship between stress and crop growth, such as growth stage and phenotype, rainfall, temperature and humidity of growth environment, etc. Therefore, it is necessary to establish a systematic stress diagnosis mechanism. Meanwhile, spectra and images are of various sources and heterogeneous data attributes, but the existing separate and independent processing approaches are difficult to obtain good performance, and the machine learning methods with simple architecture generally perform poorly in extracting high-quality features from the data with heterogeneous attributes. Additionally, for the large-scale analysis of crop stress, the instruments of spectroscopy and imaging need to be loaded on unmanned aerial vehicles and mobile vehicles, and the installation, measurement method, and data transmission and processing method should be explored.
This Research Topic aims to develop novel spectroscopy-image methods and machine learning to solve the crop stress detection problems. We welcome submissions of original research articles, reviews, and methods, including (but not limited to) research on the following sub-themes:
• Spectral and image databases for crop stress
• Fusion of spectroscopy and imaging for stress analysis
• New spectroscopy and imaging techniques
• Improvement of machine learning and deep learning
• High throughput sensing system of crop stress
• Wide-scale and mobile perception of crop stress
