Automating field based floral surveys with machine learning

The abundance and diversity of flowering plant species are important indicators of pollinator habitat quality, but traditional field-based surveying techniques are time-intensive. Therefore, they are often biased due to under-sampling and are difficult to scale. Aerial photography was collected across ten sites located in and around Rouge National Urban Park, Toronto, Canada using a consumer-grade drone. A convolutional neural network (CNN) was trained to semantically segment, or identify and categorize, pixel clusters which represent flowers in the collected aerial imagery. Specifically, flowers of the dominant taxa found in the depauperate fall flowering plant community were surveyed. This included yellow flowering Solidago spp., white Symphyotrichum ericoides/lanceolatum and purple Symphyotrichum novae-angliae. The CNN was trained using 930 m2 of manually annotated data, approximately 1% of the mapped landscape. The trained CNN was tested on 20% of the manually annotated data conceal..., Orthorectified imagery of study sites were constructed using data from a drone image acquisition program completed in the Rouge National Urban Park, Ontario, Canada during the late summer of 2021. These data represent typical late-season flowering landscapes of remnant habitat patches found in Southern Ontario, Canada. The major flowering plant groups (i.e., Solidago spp and Symphyotrichum spp) were automatically mapped using the convolutional neural netowork model trained in this study., , # Data from: Automating field based floral surveys with machine learning [https://doi.org/10.5061/dryad.nvx0k6f1t](https://doi.org/10.5061/dryad.nvx0k6f1t) This repo contains **(1)** orthorectified drone images of the ten study sites located in Rouge National Urban Park, Ontario, Canada. The imagery was collected at low altitude (7m, 15m, or 30m) in September 2024 with the DJI Phantom 4 Pro V2. **(2)** Floral classification maps predicted by the trained convolutional neural network (CNN) that is described in the paper. The CNN was trained to perform multi-classification of the three flower taxa that dominate the Fall flowering landscape in the region. **(3)** The trained CNN model that performs the multi-classification on input drone imagery. **(4)** Tabulation of plot-level floral surveys that were used to ground the truth of the CNN model. The code is provided as supporting data for Sookhan N, Sookhan S, Grewal D, MacIvor JS. 2024. Automating field-based floral surveys with machin...开花植物物种的丰度与多样性是衡量传粉者栖息地质量的重要指标，但传统的野外实地调查技术耗时耗力。因此，这类调查常因采样不足而存在偏差，且难以扩大应用规模。 本研究使用消费级无人机，在加拿大多伦多的Rouge国家城市公园（Rouge National Urban Park）及其周边共10个样地采集了航空摄影影像。我们训练了一个卷积神经网络（Convolutional Neural Network, CNN），对采集到的航空影像中代表花卉的像素簇执行语义分割任务——即识别并分类这些像素簇。具体而言，本研究针对该退化秋季开花植物群落中的优势类群花卉开展调查，涵盖开黄色花的一枝黄花属（Solidago spp.）、开白色花的毛枝帚菊（Symphyotrichum ericoides/lanceolatum）以及开紫色花的新英格兰帚菊（Symphyotrichum novae-angliae）。 该卷积神经网络使用930平方米的人工标注数据完成训练，该数据量约为研究区测绘景观总面积的1%。训练完成的卷积神经网络在20%的人工标注数据上开展了测试，相关测试内容暂未完整呈现…… 本研究构建的研究样地正射校正影像，基于2021年夏末在加拿大安大略省Rouge国家城市公园（Rouge National Urban Park）完成的无人机航摄项目数据制作。这些数据代表了加拿大安大略省南部残存栖息地斑块的典型晚季开花景观。 本研究通过训练得到的卷积神经网络模型，可自动绘制主要开花植物类群（即一枝黄花属（Solidago spp.）和帚菊属（Symphyotrichum spp.））的分布地图。 # 数据来源：Automating field based floral surveys with machine learning [https://doi.org/10.5061/dryad.nvx0k6f1t](https://doi.org/10.5061/dryad.nvx0k6f1t) 本数据集仓库包含以下四类内容： (1) 加拿大安大略省Rouge国家城市公园（Rouge National Urban Park）内10个研究样地的正射校正无人机影像。该影像于2024年9月采集，飞行高度分别为7米、15米或30米，使用设备为DJI Phantom 4 Pro V2。 (2) 本研究中训练得到的卷积神经网络（CNN）所预测的花卉分类地图，该网络被训练用于对该区域秋季开花景观中的三大优势花卉类群执行多分类任务。 (3) 可对输入的无人机影像执行多分类任务的已训练卷积神经网络模型。 (4) 用于锚定卷积神经网络模型真值的样地级花卉调查统计表。 本代码为以下论文的配套支撑数据： Sookhan N, Sookhan S, Grewal D, MacIvor JS. 2024. Automating field-based floral surveys with machine learning