论文部分内容阅读
准确提取水稻种植面积是探讨气候变化背景下水稻生产与粮食安全的重要前提。我国南方的水稻种植区域,地块破碎且受云雨天气影响严重,如何充分利用有限时相的数据获得较高精度的水稻面积提取是亟需解决的关键问题。提出了一种利用两个时相的数据,通过构建差值特征突出水稻物候变化的特点,并与随机森林算法结合高精度提取水稻种植面积的方法。将之应用于湖南省常德市鼎城区的水稻种植面积提取,结果表明:采用本方法进行水稻提取的最终总体精度达到93.01%,Kappa系数0.91,与单时相提取结果相比,总体精度提高了近3%。为了进一步分析差值特征对其他分类器的改进效果,分别将差值特征与决策树和随机森林组合,并分析了两种组合提取水稻的精度。研究发现构建的差值特征能够有效反映植物的生长状况,增加地物的可区分性,可为对象的分割及分类提供更多有用的信息,能够有效改善水稻种植面积的提取精度。
Accurate extraction of rice acreage is an important prerequisite for exploring rice production and food security in the context of climate change. How to make full use of the limited time-phase data to obtain high-precision rice area extraction is a key issue which needs to be solved urgently in the paddy planting areas and the plots in southern China are fragmented and severely affected by the rainy weather. This paper presents a method that uses the data of two phases to highlight the phenological changes of rice by constructing difference characteristics and combines the random forest algorithm with high precision extraction of rice acreage. The results showed that the final overall precision of rice extraction using this method was 93.01% and the Kappa coefficient was 0.91. Compared with the results of single-phase extraction, the overall accuracy of rice extraction was improved Nearly 3%. In order to further analyze the improvement effect of the difference feature on other classifiers, the difference features were combined with the decision tree and the random forest respectively, and the precision of the two combinations in extracting rice was analyzed. The study found that the constructed differential features can effectively reflect the growth status of plants and increase the distinguishability of features, which can provide more useful information for the segmentation and classification of objects, and can effectively improve the extraction precision of rice planting area.