Advanced imaging tools for plant phenotyping

High throughput plant phenotyping, also called plant phenomics, is an emerging and fast growing research field that aims to bridge the existing gap between genomics and plant breeding, by solving the so-called phenotyping bottleneck. Moreover, it can supply highly detailed information and tools for the advancement of both plant physiology and agronomy. Plant phenomics takes advantage from the recent developments in the fields of imaging, computer vision and sensor technologies, allowing the nondestructive detection of phenotypic characters. Plant phenomics ranges from basic science to applications in breeding and precision agriculture, combining studies performed under both controlled environments and in the open fields. Last year, Consiglio per la ricerca in agricoltura e l’analisi dell’economia agraria (CREA) joined the Italian Plant Phenotyping Network (ITA-PPN) which gathers the national research centers and universities active in this field. CREA has developed in the last ten years advanced skills for the development of analytical methods for phenotyping, mainly imaging-based. The high-throughput character of our proposed phenotyping methods should help to improve the detection of important plant traits in large field trials as well as help us to reach a better understanding of underlying yield physiological processes and facilitate the genotypephenotype associations. In particular, we developed the following analytic tools for: shape analysis 2d or 3d using landmarks (geometric morphometrics) or outline methods; quantitative color analysis from RGB images (we developed specific algorithms to standardize colors using colorchecker; 3D Thin-Plate Spline); punctual spectrophotometry and hyperspectral imaging; dynamic thermography imaging based; stereovision (multi camera systems for 3d reconstruction). Moreover, we developed an open source conveyor belt prototype multi-sensorized for rapid characterization of experimental wheat field plots. CREA developed advanced analytical approaches based on multivariate methods of prediction and classification (linear approaches and approaches based on artificial neural networks) applied in multi-parametric and multi-sensor metrology for an innovative support in phenomics. All these tools has been developed in Matlab environment, but could be easily exported in open source environments in order to realize highly customizable systems within the phenomics framework.