地表覆被作为自然过程和人类活动共同作用的重要地表景观特征,对全球或局地气候、水热循环、物质传输及陆面生态系统多样性等影响深远。利用年内时序遥感影像自动提取不同地表覆被类型的方法,以新疆阜康地区为研究目标,组织2016年植被全生长季的Landsat 8 OLI地表反射时序影像,研究不同物候期植被冠层的纹理响应信息,考察研究区典型地表覆被类型在3—11月多波段波谱、归一化植被指数（NDVI）及增强型植被指数（EVI）的时序特征,构建提取地表覆被类型的策略规则,形成时序特征匹配方法,将其应用于2018年研究区地表覆被填图的提取。最后,基于高分辨率卫星影像和野外实地调查对随机选取的2500个样点进行对比验证。结果表明：提取结果和验证数据一致性较好,总体精度为97.2%,Kappa系数为0.9655,且实地考查结果显示本方法在复播作物识别和有效降低单一时相中“异物同谱”现象上展示潜在优势。

Surface land cover is a special landscape feature modeled by natural and anthropogenic activities, significantly influencing global or local climate, hydrothermal circulation, material transport, and diversity of terrestrial ecosystem communities. Although numerous global or regional land cover datasets have been published, challenges exist in the retrieval of detailed surface cover information, especially on the automatic identification of crop types. Furthermore, the classification accuracy of land-use and land cover maps needs improvement, specifically on crops in the agricultural management community. Using temporal Landsat 8 OLI multispectral images, we analyzed the spectral features of several dominant land cover types from temporal remote sensing images, with a specific emphasis on the unique phenological rhythm of individual plants and investigated the automatic retrieval of land cover types. We organized temporal Landsat 8 OLI surface reflectance data covering Fukang City of Xinjiang, China in the 2016 vegetation growth season. According to the Fukang statistical yearbook, prior knowledge of crop phenology, and the visible texture transition of vegetation in the growing season, we systematically selected the training samples of typical surface covers. We investigated the spectrum, the normalized difference vegetation index, and the enhanced vegetation index temporal curves of 10 typical ground objects, such as agriculture fields of wheat, corn, and cotton, to extract the feature vectors modulated by the growth phase of each crop type. The Euclidean distance was calculated to measure the similarity between unknown surface targets and apriori templates built-up using the above feature vectors and assigned the land cover type according to the K-nearest neighbor rule. As a case study, we applied this approach to the land cover identification of Fukang using the time series of Landsat spectral data in 2018. We validated the retrieval accuracy using 2500 ground samples randomly collected from the land cover map of the investigated area, followed by the manual interpretation of the land cover types using the high-resolution images of 2018 and our field surveys. The results correlate well with the overall accuracy of 97.2% and the Kappa coefficient of 0.9655. Specifically, combined with the field investigation results, the method provides multi-dimensional data for ground object identification and potentially outlines the growth and development progress of different crops, enriches classification information, effectively reduces the phenomenon of foreign matters with the same spectrum in a single time phase, and has potential advantages in improving the identification accuracy of re-plant fields. The time series spectral information captured by this research well reflects the seasonal rhythm and phenological change law of crops in the Fukang region of Xinjiang and provides a novel technical method for the needs of agricultural monitoring.