Multispectral reflectance database of Cannabis sativa

Our research hypothesis postulates that different strains of Cannabis sativa grown for medicinal purposes exhibit distinctive multispectral reflectance signatures in the 410 nm to 890 nm range that can be used for nondestructive and low-cost classification. The accompanying database contains 800 spectral signatures, each with 2442 lambda reflectance values, providing detailed resolution of the optical data. These data were collected from a medical cannabis crop by randomly sampling 30 plants of two varieties and 20 plants of two other varieties, spanning three distinct phenological stages, from which both high and low leaf measurements were taken for each sample. Notable findings from this data collection will focus on identifying subtle spectral patterns and differences between varieties, as well as within-variety variability as a function of phenological stage and leaf position. The data can be interpreted to develop machine learning models capable of discriminating between medical cannabis varieties efficiently and without damaging the plant, representing a valuable alternative to current identification methods that are often destructive and costly, opening the door to applications in quality control, traceability and crop optimization in the medical cannabis industry.本研究假说认为：用于药用的不同品系大麻（Cannabis sativa）在410 nm至890 nm波段范围内具有独特的多光谱反射特征，可用于实现无损且低成本的品种分类。配套数据库包含800条光谱特征，每条包含2442个波长反射率数值，实现了光学数据的高分辨率表征。该数据集采集自药用大麻种植地块：随机选取2个品种各30株、另外2个品种各20株的植株，覆盖3个不同的物候期，且对每株采样植株分别采集上部与下部叶片的测量数据。通过本数据集可获得的重要研究结果将聚焦于：识别品种间的细微光谱模式与差异，以及由物候期和叶片位置所引发的品种内变异情况。利用该数据集可开发可高效且无损区分药用大麻品种的机器学习模型，为当前常具破坏性且成本高昂的传统鉴别方法提供了极具价值的替代方案，有望推动药用大麻产业在质量管控、溯源追踪与作物优化等领域的应用落地。