覆膜面积和曝露时长是准确评估覆膜环境效应的两个关键因素。目前覆膜信息遥感表征研究较少，且主要在像元尺度上开展单作物覆膜信息的表征，缺乏多作物覆膜信息表征方法；而且覆膜曝露时长遥感表征方法亟需研发。基于此，本项目结合面向对象影像分析方法和机器学习算法，从挖掘Sentinel-2数据对象特征优势及协同定性和定量信息的角度出发，开展多作物覆膜时空曝露信息遥感表征方法研究。首先，融入边界等特征，发展多尺度分割参数自动优化算法，获取相对最优分割精度；其次，基于随机森林回归模型提出向后剔除特征优化策略，获取最优对象特征组合，提高对象特征表征精度，获取多作物覆膜空间分布；最后，引入作物物候信息，发展基于时序物候跟踪的覆膜曝露时长遥感表征方法，获取不同覆膜曝露时长，揭示其规律，评估方法普适性。本项目在多维对象特征利用和覆膜曝露时长遥感表征方法上有创新，对农业生产监测与环境污染监测均有重要价值。

The distribution extent and exposure duration of plastic-mulching are the two key factors for accurate assessing its environmental effect. At present, there are few related studies for characterizing plastic-mulching and performed mainly at pixel scale for single crop mulching. Method for characterizing multi-crop mulching urgently in need to be developed. Moreover, there is no report on characterizing exposure duration of plastic-mulching with remote sensing. This project aims to exploit the multi-dimensional object-level features of Sentinel-2 data and collaborate qualitative and quantitative information, and develop a new method for characterizing the spatiotemporal exposure information by combining object-based image analysis and machine learning algorithm. First, an automatic algorithm for optimizing multi-scale segmentation parameter will be developed by integrating various features (such as boundaries) to obtain relative optimal segmentation accuracy. Second, a backward eliminating feature optimizing strategy will be proposed based on random forest regression model and obtain optimal object-level feature sets, and obtain the spatial pattern of multi-crop mulching using machine learning algorithm. Finally, a time-series phenology tracking method will be developed for characterizing exposure duration of multi-crop mulching by introducing the crop phenology, then reveal the exposure duration characteristic of different crop mulching, and evaluate the universality of the developed method. This project will be innovative in utilizing multi-dimensional object-level features and characterizing exposure duration of multi-crop mulching with remote sensing, will enrich theories and methods in agricultural remote sensing monitoring, and will have important value for agricultural production monitoring and environmental pollution monitoring.