High-throughput proximal ground crop phenotyping systems – A comprehensive review

Current crop phenotyping mainly relies on manual measurements and visual inspection for data collection and crop assessment, which is labor-intensive, subjective, and inefficient. Hence, modern methods depend primarily on using sensors for phenotypic data collection to replace labor vision, developing algorithms for decision-making to replace human domain knowledge, and integrating autonomous phenotyping systems to improve efficiencies in the past decades. Despite the research progress in phenotyping, there is a lack of extensive review on this topic that will be useful to various stakeholders interested in this field. Therefore, this study was conducted to perform a comprehensive review of multiple methodologies and techniques used in high-throughput ground crop phenotyping systems. A Web of Science literature search was conducted with appropriate keywords for the recent past, and the research trends in this field were captured. The current review categorizes the progress of technology in terms of phenotyping platform, sensing, data processing, and system integration. Platforms have evolved from manual-based to autonomous. Manual-based platforms require workers for data collection, while autonomous platforms involve new technologies for navigation and data collection. Different sensing techniques are used for phenotyping data collection. This study mainly discusses the mainstream sensors, including RGB, multi/hyperspectral, thermal, stereo, and light detection and ranging, and concludes that multi-source sensors could provide more accurate phenotypic information. Algorithms are applied to collected data to extract useful phenotyping information at different scales (organ, individual plant, and community). Both machine learning (ML) and deep learning (DL) have been used for phenotyping information extraction, and the DL is gradually replacing ML due to its superior performance. A case study of integrated high-throughput proximal phenotyping robot was presented, showing how different sensors and navigation systems come together to achieve on-site and real-time measurements. Advancements in high-throughput proximal ground phenotyping systems through new information, communication, sensing, and autonomous technologies in agriculture are anticipated to be more integrated and efficient phenotyping. It is anticipated that autonomous robots would finally replace workers from laborious phenotyping work.