Lootens Peter, Ruttink Tom, Rohde Antje, Combes Didier, Barre Philippe, Roldán-Ruiz Isabel
Plant Sciences Unit - Growth and Development, ILVO, Caritasstraat 39, 9090 Melle, Belgium.
Plant Sciences Unit - Growth and Development, ILVO, Caritasstraat 39, 9090 Melle, Belgium ; Bayer CropScience, Technologiepark 38, 9052 Ghent, Belgium.
Plant Methods. 2016 Jun 10;12:32. doi: 10.1186/s13007-016-0132-8. eCollection 2016.
Genetic studies and breeding of agricultural crops frequently involve phenotypic characterization of large collections of genotypes grown in field conditions. These evaluations are typically based on visual observations and manual (destructive) measurements. Robust image capture and analysis procedures that allow phenotyping large collections of genotypes in time series during developmental phases represent a clear advantage as they allow non-destructive monitoring of plant growth and performance. A L. perenne germplasm panel including wild accessions, breeding material and commercial varieties has been used to develop a low-cost, high-throughput phenotyping tool for determining plant growth based on images of individual plants during two consecutive growing seasons. Further we have determined the correlation between image analysis-based estimates of the plant's base area and the capacity to regrow after cutting, with manual counts of tiller number and measurements of leaf growth 2 weeks after cutting, respectively. When working with field-grown plants, image acquisition and image segmentation are particularly challenging as outdoor light conditions vary throughout the day and the season, and variable soil colours hamper the delineation of the object of interest in the image. Therefore we have used several segmentation methods including colour-, texture- and edge-based approaches, and factors derived after a fast Fourier transformation. The performance of the procedure developed has been analysed in terms of effectiveness across different environmental conditions and time points in the season.
The procedure developed was able to analyse correctly 77.2 % of the 24,048 top view images processed. High correlations were found between plant's base area (image analysis-based) and tiller number (manual measurement) and between regrowth after cutting (image analysis-based) and leaf growth 2 weeks after cutting (manual measurement), with r values up to 0.792 and 0.824, respectively. Nevertheless, these relations depend on the origin of the plant material (forage breeding lines, current forage varieties, current turf varieties, and wild accessions) and the period in the season.
The image-derived parameters presented here deliver reliable, objective data, complementary to the breeders' scores, and are useful for genetic studies. Furthermore, large variation was shown among genotypes for the parameters investigated.
农作物的遗传研究和育种经常涉及对田间种植的大量基因型进行表型特征分析。这些评估通常基于视觉观察和手动（破坏性）测量。强大的图像捕获和分析程序能够在发育阶段对大量基因型进行时间序列表型分析，具有明显优势，因为它们允许对植物生长和性能进行无损监测。一个包含野生种质、育种材料和商业品种的多年生黑麦草种质库已被用于开发一种低成本、高通量的表型分析工具，用于根据连续两个生长季节中单个植物的图像来确定植物生长情况。此外，我们分别确定了基于图像分析的植物基部面积估计值与刈割后再生能力之间的相关性，以及与刈割后2周的分蘖数人工计数和叶片生长测量值之间的相关性。在处理田间种植的植物时，图像采集和图像分割极具挑战性，因为室外光照条件随季节和一天中的时间而变化，且土壤颜色各异，这会妨碍在图像中勾勒出感兴趣的对象。因此，我们使用了多种分割方法，包括基于颜色、纹理和边缘的方法，以及快速傅里叶变换后得出的因子。已根据该程序在不同环境条件和季节时间点的有效性对其性能进行了分析。
所开发的程序能够正确分析处理的24048张顶视图图像中的77.2%。基于图像分析的植物基部面积与分蘖数（人工测量）之间以及刈割后再生（基于图像分析）与刈割后2周的叶片生长（人工测量）之间发现了高度相关性，r值分别高达0.792和0.824。然而，这些关系取决于植物材料的来源（饲料育种系、当前饲料品种、当前草坪品种和野生种质）以及季节中的时期。
此处呈现的基于图像的参数提供了可靠、客观的数据，是对育种者评分的补充，对遗传研究很有用。此外，在所研究的参数方面，基因型之间表现出很大差异。
