Toward digital and image-based phenotyping

Abstract

Manually performed measurements in field phenotyping are labor- and time-consuming, often destructive and not objective. Moreover, the complexity and variability of crop plants under field conditions require highresolution data and filtering. As a consequence, there is a need for spatially and temporally differentiated data, objective data acquisition, and high-throughput technologies. Image-based systems, selective on morphological and spectral crop characteristics, are adequate sensors for further interpretation of raw data in terms of crop properties. In particular multi-sensor and data fusion has a potential to compensate the varying influences of sunlight, dust, moisture, or uneven land in the field. Due to the high-resolution data of image-based systems-such as digital color cameras, spectral imaging, laser scanning devices, or 3D cameras-detailed crop properties have become available, even individual plant phenotyping is an option. Autonomous field robots have a high potential for field phenotyping as well as new sensor technologies and virtual phenotyping. Data management is of relevance for field phenotyping, starting from storing the large amounts of raw data up to artificial intelligence algorithms for trait determination. Interdisciplinary cooperation is crucial for the implementation of digital phenotyping into practice.