Click on “Download PDF” for the PDF version or on the title for the HTML version.

If you are not an ASABE member or if your employer has not arranged for access to the full-text, Click here for options.

Abstract. The development of new automated technologies for high-throughput in-field phenotyping of plant breeding candidates are urgently needed by plant breeders to achieve rapid development of new crops varieties to provide food security for a growing population under the threat of climate change. Phenotyping in agriculture is often done by hand and can utilize destructive methods, making it the bottleneck in breeding climate resistant crops. In-field phenotyping is used to characterize how candidate breeding material performs in a real environment, something that can‘t be done in a greenhouse. High-throughput phenotyping methods need to be both cost effective and accurate. In biological scenes, traditional machine vision techniques like color segmentation and shape recognition are often ineffective due to the complexity of the scenes and the occlusion of objects. In tomatoes, finding visible red fruit in a group of green leaves is straightforward using classical machine vision techniques, but finding green fruit among green leaves can be challenging because their color is similar, making it difficult to quickly find them.

Our engineering group at UC Davis has developed a new technique to distinguishing green tomato fruit from tomato leaves for high-throughput in-field phenotyping applications. This technique uses visible and near infrared reflectance techniques to create a high-contrast pseudo color representation of the outdoor scene. A digital single-lens reflex camera was modified to sense near-infrared light in addition to visible light for this application. Using this method, a machine vision index was created that amplifies the visual differences between green fruit and green leaves without compromising the ability to use shape recognition or other visual features of the scene besides color. This paper reports on the development and performance of this novel high-throughput phenotyping method for characterizing green fruits in vegetable crops.