Use of UAV images for precision agriculture and environmental applications

对于一些环境应用，机载/卫星图像不如无人驾驶飞行器（UAV）图像合适。无人机可以在非常低的高度操作，从而产生高空间分辨率的图像。它们更具移动性、速度更快、适应性更强，而且更易于使用。此外，它们允许以重复和更便宜的方式进行调查。在NSERC-CRD的资助下，我们正在开发一种用于作物管理的新型多光谱无人机相机。所请求的发现拨款将建立在NSERC-CRD拨款和几个正在进行的项目的基础上。首先，我们将改进我们开发的从无人机图像中自动计数果园树木的方法，以检测受损树木。我们还将在其他类型的果园试验这种方法。其次，我们将测试使用无人机图像来监测大叶草床的恢复，并结合加拿大环境部的一个成功的赠款支持现场和无人机数据采集。第三，无人机图像将用于监测最近恢复的盐沼上的自然植被再定植。对于这三个项目，将开发利用光谱和纹理图像信息的新方法。考虑到图像的高空间分辨率，除了标准的基于像素的分类器之外，还将测试高级分类器。无人机图像处理可能具有挑战性，因为无人机更容易受到大风条件的影响，这可能会影响观察和照明几何形状。此外，无人机的低飞行高度需要更多的图像来覆盖该区域，这可能导致传感器温度的增加和变化天气条件的概率。对于大叶藻项目，另一个挑战是考虑水柱对图像采集和处理的影响。此外，对于这两种沿海应用，图像采集应考虑到潜在的太阳闪烁效应。这项研究对那些需要一种负担得起的、灵活的监测工具来提供高空间图像的人来说很重要。受损树的测绘方法将被种植者用来评估果园的状况，并被作物保险机构用来确定对种植者的赔偿。渔业和海洋部和其他从事大叶藻恢复项目的非政府组织将使用这种测绘方法。加拿大鸭子无限公司将使用盐沼测绘方法来监测盐沼恢复情况。除了这三种应用之外，该研究还将产生可用于其他应用的无人机图像处理的新方法。拟议的研究将有助于加拿大蓬勃发展的商用无人机领域的发展，预计未来五年全球销量将增长四倍以上。

For several environmental applications, airborne /satellite images are less suitable than unmanned aerial vehicles (UAV) images. UAVs can operate at much low altitudes that results in high spatial resolution images. They are more mobile, fast, adaptable, and easier to use. Furthermore, they allow surveys in a repetitive and much more cheaply way. Under a NSERC-CRD grant, we are developing a new multispectral UAV camera for crop management. The requested Discovery Grant will built upon this NSERC-CRD grant and several on-going projects. First, we will improve the automated method we developed to count apple orchard trees from UAV images to detect damaged trees. We will also test the method to other types of orchards. Second, we will test the use of UAV images for monitoring eelgrass bed restoration, in conjunction to a successful Environment Canada grant supporting field and UAV data acquisition. Third, UAV images will be used to monitor natural vegetation recolonization on recently restored salt marshes. For the three projects, novel methods that use spectral and textural image information will be developed. Given the high spatial resolution of the images, advanced classifiers will be tested in addition to standard pixel-based classifiers. UAV image processing can be challenging because UAV are more susceptible to windy conditions, which can affect the viewing and illumination geometries. Also, the low UAV flying altitude requires more images to cover the area, which can lead to increasing of the sensor temperature and of the probability of variable weather conditions. For the eelgrass project, an additional challenge is to consider the effect of the water column on the image acquisition and processing. Also for both coastal applications, image acquisition should take into account potential sun glint effects. The research is important to those who need an affordable and flexible monitoring tool that provides high spatial images. The damaged tree mapping method will be used by growers to assess the status of their orchards and by crop insurance agencies to determine compensations to growers. The eelgrass mapping method will be used by the Department of Fisheries and Oceans and other NGO groups working on eelgrass restoration projects. The salt marsh mapping method will be used by Ducks Unlimited Canada to monitor salt marsh restoration. Beyond these three applications, the research will also produce novel methods for UAV image processing that can be used in other applications. The proposed research will contribute to the development in Canada of a booming commercial UAV sector that is projected to more than quadruple in sales worldwide over the next five years.