Research On Parallel Method Faced To Digital Terrain Analysis With Data Dependence

With the rapid development of the parallel computing technology, the research of parallel digital terrain analysis has gradually become the focus of the digital terrain analysis in the background of the explosive growth of spatial data. If the traditional serial algorithm of digital terrain analysis is simply parallel, it can’t get the corresponding parallel efficiency because of the influence of some factors, such as the dependence of the internal calculation and some factors of the parallel computing platform. Therefore it has important significance to study how to make the algorithm of digital terrain analysis parallel which has data dependence. Meanwhile, the optimize analysis of the parallel digital terrain analysis algorithm in parallel computing platform can further improve the efficiency.Firstly, by analyzing the calculation of some different terrain factors in some algorithms of digital terrain analysis, the data dependence of the parallel algorithm is summed up. According to the different data dependence of the local digital terrain analysis and the global digital terrain analysis, a hierarchical scheduling model facing to the data dependence is put forward by analyzing the parallel strategy of parallel algorithms of digital terrain analysis. Then the applicability of the model is analyzed.Secondly, based on the theory of the hierarchical scheduling model, the application of some specific algorithms of digital terrain analysis is respectively analyzed. For example, the model is introduced into the parallel viewshed analysis algorithm, the parallel watershed analysis algorithm with data dependence and the parallel shortest path algorithm. Then the above application is further researched.Moreover, the characteristic of data reading and distribution of the parallel digital terrain analysis algorithm can be analyzed through the analysis of the efficiency of I/O and the node communication in the distributed cluster environment. Then an optimization strategy of data reading and distribution is presented which can be applied to the parallel algorithm of digital terrain analysis. Through the optimization strategy, it can optimize the reading, sending, writing process of the parallel digital terrain analysis algorithm in the distributed cluster environment.Finally, in the cluster system, to ensure the performance of the data redundant partition strategy for the local digital terrain analysis, the experiment is testing by different experimental data. According to results, the strategy can realize the parallel process of the local digital terrain analysis. Meanwhile the experiment is also realized to test the performance of the optimization strategy of the hierarchical scheduling model based on the parallel shortest path algorithm. Then the optimization strategy of data reading and distribution is applied to the maximum slop algorithm. By analyzing the results, the optimization strategy can further improve the performance of the parallel algorithm of digital terrain analysis.