IIS: EAGER: Benchmarks for Autonomous Unmanned Aerial Vehicles in Agriculture Applications

可以开发无人机技术和航空图像，以帮助农民在不使用更多土地的情况下满足未来提高作物产量和降低成本的需求。然而，获得这项技术、进入田野以及获得飞行和捕捉图像的许可可能会限制开发人员的数量。该项目设计了基准，以捕获、表征和共享从部署在农业环境中的自主无人机(AUAV)系统收集的经验数据。它管理和共享俄亥俄州TDA数据共享中心的AUAV数据，这将使研究能够跨学科进行，并将在云研讨会上主持农业分析，吸引计算机科学家、地球科学家、农民和农业工程师。一个具有启发性的应用例子是作物间伐任务，它通过选择性地对杂草和脆弱的作物应用除草剂来防止植物之间的竞争。人工间伐对体力要求很高，每英亩的成本可能高达100美元。用于作物间伐的自主无人机系统将需要能够处理图像，以识别各种植物类型的拥挤。如果检测到过度拥挤，AUAV可以降低并悬停以捕获适合于强作物、弱作物和杂草分类的高分辨率图像。该项目创建了一个AUAV系统的参考实现，该系统可以检测作物间伐，同时捕获低分辨率和高分辨率图像。研究人员可以模仿这种AUAV系统，用不同的硬件和软件进行整体测试，并使用这些基准数据来探索在不需要飞行AUAV的农田的情况下，对拥挤和作物强度进行机上或在线分类的方法。还可以记录Auav功率，以帮助探索充电和采样较少数据的决策。

Drone technology and aerial imagery can be developed to help farmers meet future demands to increase crop yield and lower costs without using more land. Access to the technology, to the fields, and to permits to fly and capture imagery can limit the pool of developers, however. This project devises benchmarks that capture, characterize and share empirical data collected from autonomous unmanned aerial vehicle (AUAV) systems deployed in agricultural settings. It curates and shares AUAV data on the Ohio State TDA Data Commons which will enable research across disciplines, and will host an Agriculture Analysis in the Cloud workshop which attracts computer scientists, geoscientists, farmers and agricultural engineers.A revealing example of application is the task of crop thinning, which prevents competition between plants by applying herbicide selectively to weeds and weak crops. Manual thinning is physically demanding and can cost up to $100 per acre. An autonomous UAV system for crop thinning will need to be able to process imagery to identify crowding in various plant types. If over-crowding is detected, the AUAV can lower and hover to capture hi-res images suitable for classification of strong crops, weak crops and weeds. This project creates a reference implementation of an AUAV systems that detects crop thinning, capturing both low and high resolution imagery. Researchers can mimic this AUAV system to perform holistic tests with different hardware and software, and use this benchmarked data to explore approaches for on-board or on-line classification of crowding and crop strength without access to farmland for flying AUAVs. AUAV power can also be recorded, to help explore decisions on charging vs sampling less data.