本文研究区位于华北地区保定市，属于黄淮海农业气候区域，实行“冬小麦-夏玉米”一年两熟种植制度。在夏玉米生长发育期间，保定地区多种植蔬菜、棉花、大豆、辣椒等经济作物以及果树，容易与夏玉米产生混淆，同时农田周边分散的杂草也是识别和提取夏玉米的阻碍。针对这一现状，本文提出了以时序NDVI为特征量的使用不同分辨率遥感数据的保定地区夏玉米的识别和提取方法：

（1）Meta分析方法。对有关作物NDVI值的关键词进行检索，收集研究区附近地区记录或包含作物NDVI值的文献。分析夏玉米和其他同时段生长的作物在不同生育期NDVI值的差异。对研究区域范围内的作物在不同生育期内的NDVI特征值进行Meta分析，找到夏玉米与其他作物的NDVI值存在显著差异的生育期。确定能够通过不同生育期的夏玉米与其他作物的NDVI特征值来提取作物，获取夏玉米与其他作物区分的关键生育期和对应的NDVI阈值。并通过分析结果在遥感影像上选择夏玉米和非夏玉米样本。

（2）夏玉米种植区提取。对Landsat 8数据、哨兵二号数据、高分二号数据进行预处理，并得到时序NDVI曲线。在研究区域内选取夏玉米和其他典型作物，制作时序NDVI曲线，说明遥感数据在本地区的适用性和与Meta分析结果的一致性。通过Meta分析结果选择的夏玉米与非夏玉米的样本以及像元的NDVI特征值，使用支持向量机算法、梯度提升决策树算法、随机森林算法三种分类器，分别提取夏玉米种植区。对不同数据和方法的提取结果作出比较和分析。

基于时序NDVI的夏玉米的识别和提取相对于其他的作物提取方法，精度较高，在区分和夏玉米生长状况较为接近的作物的时候有较好的效果，总体精度在0.8以上，提取的夏玉米种植区域占保定市面积与实际值12.8%接近。从种植结构方面看，此方法对于夏玉米大面积种植区域、种植结构较为复杂的地区以及山林与耕地混杂地区，都有较好的提取结果；从分辨率方面看，在高分辨率的遥感影像中提取作物的精度比低分辨率的遥感影像要高，尤其是在保定市种植结构较为复杂的地区。

通过本文的研究，发现夏玉米与其他作物在不同生育期的NDVI特征值差异较大，可以据此提取夏玉米种植区域。提取结果显示，保定市夏玉米主要分布在东南部平原地区，西北部丘陵山地种植较少。

The study area in this paper is located in Baoding City, North China, which belongs to the Huanghuaihai Agricultural Climate Zone. The "winter wheat-summer maize " planting system is implemented in a year. During the growth and development of summer maize, Baoding also grows more economic crops such as vegetables, cotton, soybeans, peppers, and fruit trees. At the same time, weeds scattered around the farmland are also obstacles to identifying and extracting summer maize. In response to this situation, this paper proposes a method for identifying and extracting summer maize in Baoding using remote sensing data of different resolutions using time series NDVI as the characteristic quantity:

(1) Meta-analysis method. Search for keywords related to crop NDVI values, and collect records in the vicinity of the study area or documents containing crop NDVI values. Analyze the difference in NDVI value of summer maize and other crops grown at the same time in different growth periods. Meta-analysis of the NDVI characteristic values of the crops in the study area in different growth periods is carried out, and the growth periods in which the NDVI values of summer maize and other crops are significantly different are found. It is determined that crops can be extracted from the NDVI characteristic values of summer maize and other crops in different growth periods, and the key growth periods and corresponding NDVI thresholds for distinguishing summer maize from other crops can be obtained. And through the analysis results, select summer maize and non-summer maize samples on the remote sensing image.

(2) Extraction from summer maize planting areas. The Landsat 8 data, Sentinel-2 data, and GF-2 data are preprocessed, and time series NDVI curves are obtained. Select summer maize and other typical crops in the study area, and make time series NDVI curves to illustrate the applicability of remote sensing data in this area and the consistency with the results of Meta-analysis. The samples of summer maize and non-summer maize selected through the results of the Meta-analysis and the NDVI feature values of the pixels are used to extract the summer maize planting areas using three classifiers: support vector machine algorithm, gradient boosting decision tree algorithm, and random forest algorithm. Compare and analyze the extraction results of different data and methods.

Compared with other crop extraction methods, the recognition and extraction of summer maize based on time series NDVI has higher accuracy, and has better results in distinguishing crops that are closer to the growth status of summer maize. The overall accuracy is above 0.8. The maize planting area accounts for 12.8% of the actual value of Baoding City. From the perspective of planting structure, this method has good extraction results for large-scale summer maize planting areas, areas with more complex planting structures, and areas with mixed forests and arable land; from the perspective of resolution, it can be used in high-resolution remote sensing images. The accuracy of extracting crops in the medium is higher than that of low-resolution remote sensing images, especially in areas where the planting structure is more complex in Baoding.

Through the research in this article, it is found that the NDVI characteristic values of summer maize and other crops are quite different in different growth periods, and the summer maize planting area can be extracted accordingly. The extraction results show that summer maize in Baoding City is mainly distributed in the southeastern plains, while the hills and mountains in the northwest are less planted.