利用面向对象变化向量分析(OCVA)检测土地利用变化

Land use change detection based on object-oriented change vector analysis (OCVA)

关键词:  土地利用  变化向量分析  面向对象  高分二号  欧氏距离

Key words:  land use  change vector analysis  object-oriented  GF-2  Euclidean distance

摘要:

为推广国产高分数据在土地利用变化检测方面的应用，以延庆区张山营镇2015和2018年2期高分二号（GF-2）影像为数据源，采用面向对象变化向量分析法（OCVA），先通过不同分割模式获取对象，其次利用引入权重的欧氏距离构建变化向量的模，再通过目标函数确定最佳检测阈值后对研究区进行变化检测，最后对变化区域进行面向对象分类，得到具体的"从…到"变化类型。结果表明：1）多时相组合分割模式下的OCVA法精度要高于多时相分别分割模式；2）利用加权组合方法构建对象变化向量的模更易确定最佳检测阈值，且最后的总体精度要高于传统OCVA法；3）结合面向对象分类的OCVA法能有效减少面向对象分类后变化检测法和单一OCVA法的伪变化，检测精度达到92.50%。

Abstract:

To promote the application of domestic Gaofen data in land-use change detection,taking the two GF-2 images in 2015 and 2018 of Zhangshanying Town as the data source,the object-oriented change vector analysis(OCVA) was adopted in this study.First,the objects were obtained by different segmentation modes,and the Euclidean distance of the weight was used to build the mold of change vector.Second,the optimal detection threshold which was determined by the objective function was used to detect the whole study area.Finally,the "from-to" change type was obtained by the object-oriented classification of the change region.The results showed that:1) The accuracy of OCVA method in multi-temporal combined segmentation mode was higher than that in segmentation mode;2) The optimal detection threshold was easier to determine through using weighted combination method to build the mold of change vector,and its overall accuracy was higher than the traditional OCVA method;3) Compared with the object-oriented post classification change detection and the single OCVA method,the OCVA method combined with object-oriented classification could effectively reduce the pseudo-change,and the detection accuracy of which reached 92.50%.