多波束测深数据并行滤波算法

Parallel filtering algorithm for multibeam bathymetry data

针对提高节结合不确定度的水深评估算法执行速度的问题,该文在传统的CUBE滤波算法基础上提出一种适用于多核架构的并行滤波算法。把测深数据滤波分解为构造CUBE网格和估算网格节点水深两个环节,然后针对各环节特点设计对应的并行计算方法。通过两种方法提高滤波速度:一是通过合理的并行策略把计算任务分配到多个处理器核上并行执行;二是引入基于堆结构的排序方法提高数据滤波速度。实验结果表明,所设计算法在四核处理器上运行时,能够将滤波速度提高约2.74倍。

In order to improve the processing speed of multibeam bathymetry data filtering algorithm, a multicore parallel filtering algorithm was designed based on the traditional combined uncertainty and bathymetry estimator （CUBE） algorithm. In this algorithm, the bathymetry data filtering procedure was decomposed into two steps： first CUBE grid was constructed and the depth of each grid node was estimated, then different parallelism was designed according to the character of each step. This algorithm introduced two speedup methods： the first one was assigning the computing task to CPU cores by elaborate parallelism strategy; the second one was improving the data filtering speed by heap sort algorithm. Experiment results showed that this algorithm could increase the processing speed by about 2.74 times in a quad-core platform.