Height is a key trait in the indices applied when selecting genotypes for use in both tree breeding populations and production populations in seed orchards. Thus, measurement of tree height is an important activity in the Swedish Norway spruce breeding program. However, traditional measurement techniques are time-consuming, expensive, and often involve work in bad weather, so automatization of the data acquisition would be beneficial. Possibilities for such automatization have been opened by advances in unmanned aerial vehicle (UAV) technology. Therefore, to test its applicability in breeding programs, images acquired by a consumer-level UAV (DJI Phantom 4 Pro V2.0) system were used to predict the height and breast height diameter of Norway spruce trees in a 12-year-old genetic field trial established with 2.0 × 2.0 m initial spacing. The tree heights were also measured in the field using an ultrasonic system. Three additive regression models with different numbers of predictor variables were used to estimate heights of individual trees. On stand level, the average height estimate derived from UAV data was 2% higher than the field-measured average. The estimation of family means was very accurate, but the genotype-level accuracy, which is crucial for selection in the Norway spruce breeding program, was not high enough. There was just ca. 60% matching of genotypes in groups selected using actual and estimated heights. In addition, heritability values calculated from the predicted values were underestimated and overestimated for height and diameter, respectively, with deviations from measurement-based estimates ranging between −19% and +12%. However, the use of more sophisticated UAV and camera equipment could significantly improve the results and enable automatic individual tree detection.

中文翻译：

---

使用无人机进行表型评估后的12年挪威克隆云杉田间试验遗传参数估计和优良基因型选择

在选择用于种子园的树木育种种群和生产种群的基因型时，高度是所应用指标的关键特征。因此，测量树高是瑞典挪威云杉育种计划中的一项重要活动。但是，传统的测量技术非常耗时，昂贵，并且通常涉及恶劣天气下的工作，因此自动进行数据采集将是有益的。随着无人机技术的发展，这种自动化的可能性已经打开。因此，为了测试其在育种程序中的适用性，使用了消费者级无人机（DJI Phantom 4 Pro V2.0）系统获取的图像来预测挪威云杉树的高度和胸高直径，该云杉树具有12年的遗传初始间距为2.0×2.0 m的现场试验。还使用超声系统在野外测量了树木的高度。使用具有不同数量的预测变量的三个加性回归模型来估计单个树木的高度。在看台水平上，从无人机数据得出的平均高度估计值比现场测量的平均值高2％。对家庭平均数的估计非常准确，但是在挪威云杉育种计划中选择至关重要的基因型水平准确性还不够高。大约只有 使用实际和估计身高选择的组中，基因型的匹配率达到60％。此外，分别根据高度和直径低估和高估了从预测值计算得出的遗传力值，与基于测量的估计值之间的偏差介于-19％和+ 12％之间。然而，