Directional solidification experiments of transparent alloy systems typically show multiple dendrites, a forest of dendrites, growing with preferential alignment. At the length scale of centimetres, an experiment could have hundreds of observable dendrites. Analysis of every dendrite would be laborious and practically difficult to implement. Hence, low numbers of dendrites are routinely selected for analysis as they are assumed to be representative of the growth conditions. Hence, many dendrites go without being analysed. Here, a bespoke experimental apparatus with a novel computer vision algorithm is presented that automatically detects and simultaneously tracks multiple columnar dendrite tips from in-situ video data of directional solidification. The benefits of the algorithm are demonstrated with an application to an experimental test case with the transparent alloy system Neopentyl Glycol-35 wt%D-Camphor (NPG-35 wt%DC). Comparisons of dendrite tip velocity and undercooling measurements with microgravity experimental results from the literature showed notable differences. The current terrestrial data showed similar growth rates but at lower undercoolings (by factors in the range of 0.41–0.68) to that measured in the microgravity experiments. Comparisons were made to the classical Lipton-Glicksman-Kurz (LGK) model and to a modified LGK model adapted with a finite diffusional boundary layer theory to account for convection effects. The modified LGK model showed good agreement for boundary layers between 2.5 and 7.0 µm. An oscillatory component to the tip velocity was observed between adjacent columnar dendrites. Video data of columnar dendritic growth augmented with tip velocity vectors are presented. The tip tracking algorithm is beneficial as, with 385 dendrite tips tracked, it provides statistical and qualitative insights that are otherwise difficult to reconcile using traditional methods.

透明合金系统的定向凝固实验通常显示多个枝晶，枝晶森林，以优先排列的方式生长。在厘米的长度尺度上，一个实验可能有数百个可观察的树突。对每个枝晶的分析将是费力的并且实际上难以实施。因此，通常选择少量的树突进行分析，因为它们被假定为代表生长条件。因此，许多树突没有被分析。在这里，提出了一种具有新颖计算机视觉算法的定制实验装置，该装置可以从定向凝固的原位视频数据中自动检测并同时跟踪多个柱状枝晶尖端。D-樟脑 (NPG-35 wt% DC)。枝晶尖端速度和过冷度测量与文献中的微重力实验结果的比较显示出显着差异。当前的地面数据显示出与微重力实验中测量的增长率相似的增长率，但过冷度较低（系数范围为 0.41-0.68）。对经典的 Lipton-Glicksman-Kurz (LGK) 模型和修改后的 LGK 模型进行了比较，该模型采用有限扩散边界层理论来解释对流效应。修改后的 LGK 模型对 2.5 和 7.0 µm 之间的边界层显示出良好的一致性。在相邻柱状树突之间观察到尖端速度的振荡分量。呈现了用尖端速度矢量增强的柱状树突生长的视频数据。