低质量遥感图像的时空两维联合挖掘实现仿真

Low Quality Image Two-Dimensional Combined Mining to Realize the Simulation of Space and Time

研究遥感图像的准确挖掘定位问题。在遥感图像采集中,大量的低质量图像会使真正需要定位图像的特征变模糊。传统的基于决策树的图像挖掘分类器,无法处理低质量图像中存在的数据缺失以及噪声干扰,导致图像数据库中某些属性的字段缺值或数据不完整或丢失,使产生的决策树无法对图像数据进行有效的挖掘分析。提出一种时空两维联合的低质量图像挖掘方法。首先,在Apriori算法中融入时间维度以及空间维度。然后,在不同的时间段以及空间层中,通过最小能耗监控单元,塑造分时分层的数据挖掘模型,挖掘出更多的图像关联知识。最后,采用自适应遗传方法对算法的挖掘过程进行优化,避免算法出现局部最佳解。仿真实验结果表明,改进方法的挖掘时间开销以及空间开销都优于决策树方法,能够对低质量遥感图像进行快速、准确的挖掘,具有较高的应用价值。

In this paper, the accuracy of remote sensing image mining and locating was researched. In remote sensing image acquisition, a large number of low-quality images will blur the features of images really need to be lo- cated. This paper presented low-quality images combined mining method based on a two-dimensional space-time. First, the time dimension and spatial dimensions were integrated into the Apriori algorithm. Then, in a different time period and spatial layer, by minimizing energy consumption monitoring unit, we shaped timeshare and hierarchical data mining model, and dug out more images related knowledge. Finally, an adaptive genetic method was utilized to optimize the mining process of the algorithm to avoid the partial optimal solution. Simulation results show that the ex- cavation time overhead and space overhead of the proposed method are superior to the decision tree method, capable of low-quality remote sensing images for fast and accurate excavation, and has a high application value.