NCEA's Remote Monitoring and Automatic Detection of Grain Crop Attributes for GRDC Variety Trials

The National Variety Trials (NVT) involve a yearly coordination of 630 grain trials conducted across 250 locations in Australia. At different stages of the crop season Trial Service Providers visually assess the attributes of the grain plants in each trial-plot to evaluate the growth and development of the different grain varieties. This involves manual measurements related to: (i) plant dimensions (height, canopy size); (ii) different stages of growth (seedling, tillering, jointing, boot and flowering); and (iii) germination rate. However, the availability of personnel to perform this monitoring is likely to be constrained to larger research stations. These plant attributes can be visually monitored and automatically detected using remote camera-based machine vision technologies to improve the timeliness and consistency of assessment of the grain varieties. In addition to streaming visual data of the crop, there is potential for machine vision technology to automatically analyse the images to determine a range of plant attributes and performance indicators from video-frame samples collected daily; such as flowering behaviour (50% of the plot to anthesis) and crop height. The data captured, processes and stored will be used to determine variation between varieties of grains across Australia.国家品种试验（National Variety Trials，NVT）每年会在澳大利亚250个试验点协调开展630项谷物试验。在作物生长季的不同阶段，试验服务提供商将对每个试验小区内的谷物植株性状开展目视评估，以评价不同谷物品种的生长发育状态。该评估涉及以下人工测量内容：(i) 植株性状（株高、冠幅）；(ii) 不同生长阶段（苗期、分蘖期、拔节期、孕穗期及开花期）；(iii) 发芽率。然而，可承担此类监测工作的人员资源往往仅能覆盖规模较大的科研站点。通过采用基于远程摄像的机器视觉技术，可实现植株性状的目视监测与自动识别，进而提升谷物品种评估的时效性与一致性。除采集作物的实时视觉数据流外，机器视觉技术还可通过自动分析每日采集的视频帧样本，测定一系列植株性状与性能指标，例如开花特性（试验小区50%植株进入扬花期）及作物株高。所采集、处理与存储的全部数据，将用于分析澳大利亚境内不同谷物品种间的性状差异。