Plant disease severity estimated visually: a century of research, best practices and opportunities for improving methods and practices to maximize accuracy

Abstract

Plant disease quantification, mainly the intensity of disease symptoms on individual units (severity) is the basis for a plethora of research and applied purposes in plant pathology and related disciplines. These include evaluating treatment effect, monitoring epidemics, understanding yield loss, and phenotyping for host resistance. Although sensor technology has been available to measure disease severity using the visible spectrum or other spectral range imaging, it is visual sensing and perception that still dominates, especially in field research. Awareness of the importance of accuracy of visual estimates of severity began in 1892, when Cobb developed a set of diagrams as an aid to guide estimates of rust severity in wheat. Since that time, various approaches, some of them based on principles of psychophysics, have provided a foundation to understand sources of error during the estimation process as well as to develop different disease scales and disease-specific illustrations indicating the diseased area on specimens, similar to that developed by Cobb, and known as standard area diagrams (SADs). Several rater-related (experience, inherent ability, training) and technology-related (instruction, scales and SADs) characteristics have been shown to affect accuracy. This review provides a historical perspective of visual severity assessment, accounting for concepts, tools, changing paradigms, and methods to maximize accuracy of estimates. A list of best operating practices in plant disease quantification and future research on the topic is presented based on the current knowledge.

Identification of resistance for Phyllachora maydis of maize in exotic‐derived germplasm

“…An ordinal scale was developed to evaluate plots for resistance to P. maydis (Table 3). While the use of a standard area diagram has been viewed as a more accurate method of disease evaluation (Bock et al., 2021), the ordinal scale employed here allows for rapid whole plot evaluation of large numbers of genetically distinct plots. Previous studies for tar spot where large numbers of plots were evaluated for resistance also utilized an ordinal scale and successfully identified genetic variation associated with resistance (Cao et al., 2017; Mahuku et al., 2016).…”

Attention-optimized DeepLab V3 + for automatic estimation of cucumber disease severity

“…The R 2 of different backbones are generally higher than 0.87, and the RMSE values are lower than 2.7, except for Xception. The results show that the severity estimation models based on pixel-wise classification can reasonably estimate the disease severity [3]. Moreover, most disease severity is overestimated when estimating the disease severity by the segmentation results of the semantic segmentation models (Fig.…”

“…The accuracy for the five severity categories that were defined in this method was 86.51%. Although accurate results were reported in the above studies, dividing severity percentages into multiple categories in field trials did not make it easy to assess the effectiveness of treatments, such as fungicides [3].…”

“…Visual estimation is the conventional approach to quantify the disease severity, which assign a severity value to the symptoms perceived by the human eye in the visible light range. Disease severity based on ratio scales is usually performed by manual estimation of the visual score according to the number and the area of the plants' lesions [3]. It has been proved the most accurate tool for estimating the disease severity [4].…”

Agricultural Plant Cataloging and Establishment of a Data Framework from UAV-based Crop Images by Computer Vision

“…Such data can be used to extract phenotypic traits such as diameter or height of cauliflower plants or their head [8,13]. Additionally, plant disease research in general is an excellent use case for this workflow since it requires having individual plant information throughout many growth stages [14,15,16]. Thus, a spatio-temporal individualization of the investigated plants is crucial.…”