支持向量机与分类后验概率空间变化向量分析法相结合的冬小麦种植面积测量方法

利用遥感手段提取农作物种植面积时，需要结合作物物候特征，以提高面积的提取精度。该文以北京市通州区西南部为试验区，以冬小麦为研究对象，利用多时相的环境减灾小卫星遥感影像数据，通过基于支持向量机二分法的分类后验概率空间变化向量分析法进行冬小麦种植面积遥感测量试验研究。研究结果表明：该文提出的方法测量结果总体精度、Kappa系数分别为95%、0.90，远高于支持向量机（SVM）分类后直接比较方法（总体精度91%，Kappa系数0.79）；解决了实际应用中的变化阈值选取的主观性问题，该方法的频度直方图两极化现象使得变化阈值取值部分频度被压低摊平，阈值敏感度降低，变化阈值取值更为客观，一定程度上解决了阈值难以设定的问题；SVM二分法和变化向量分析的结合增强了对光谱的敏感性，能够监测不同季相上植被的长势变化，进而提高了农作物种植面积遥感测量的精度，同时对其他农作物种植面积测量提供了途径。

Method of winter wheat planting area estimation based on support vector machine and post-classification changed vector analysis

The crop phenology characristics can greatly improve estimation of planted area while using remote sensed technologies. Taking Southeast Beijing as the study area in this paper, the support vector machine (SVM) dichotomy model and post-classification changed vector analysis (PCVA) model were integrated to estimate winter wheat area. The results indicate that as follows: The overall pixel accuracy and Kappa coefficient resulted from this proposed method were 95% and 0.90, which were much better than those from post-classification comparison method (91% and 0.79). The combining of SVM and PCVA models also presented a good help on the selection of changing threshold value which tended to be subjective. Besides, with the polarization phenomenon of the frequency histogram in this method, it decreased the partial frequency of change threshold value and led to a lower threshold sensitivity, thus the determination of threshold value was more objective. The combining use of SVM and PCVA models was more sensitive to spectral changes, and improved the detection of crop growth change under different growing stages, as well as the estimating accuracy on winter wheat planted area. It is believed that this method also has a great potential for other crops planted area estimates.