Early season crop classification using multi-frequency polarimetric synthetic aperture radar with machine learning methods

及时和准确的作物分类对加拿大的作物管理和粮食安全至关重要。目前，加拿大作物分类地图仅在每个生长季结束时提供，因此早期作物分类将有助于加拿大更好地管理农业活动。大多数作物都有类似的光谱特征，生长早期频繁的多云天气对光学卫星图像来说是一个挑战。当前合成孔径雷达(SAR)图像的应用受到重复周期较长的限制，因此利用多频率、多角度的SAR图像是实现准确的早季作物分类的关键。该项目的目标是扩展A.U.G.Signals Ltd.(AUG)的作物生长阶段估计技术，利用Sentinel-1、TerraSAR-X和模拟RADARSAT星座任务(RCM)极化SAR(PolSAR)图像，结合最先进的机器学习算法，实现早季作物分类。该项目的成果将帮助AUG将作物监测技术商业化，为加拿大农业提供及时和准确的信息。

Timely and accurate crop classification is essential for Canada's crop management and food security. Currently, Canadian crop classification maps are only available at the end of each growing season, therefore early season crop classification will help Canada better manage agricultural activities. Most crops have similar spectral signatures and frequent cloudy days in early growing seasons are challenges for optical satellite images. The application of current synthetic aperture radar (SAR) images is limited by longer repeating cycles so the utilization of multi-frequency and multi-angle SAR images are essential for achieving accurate early season crop classification. The objective of this project is to extend the crop growth stage estimation technology of A.U.G. Signals Ltd. (AUG) to achieve early season crop classification using Sentinel-1, TerraSAR-X and simulated RADARSAT constellation mission (RCM) polarimetric SAR (PolSAR) images with state-of-the-art machine learning algorithms. The outcome of the project will help AUG commercialize crop monitoring technologies to provide timely and accurate information for the Canadian agriculture industry.