顾及空间邻接关系的无人机影像匹配并行算法

发布时间：2018-07-23 16:09

【摘要】：为解决无人机(unmaned aerial vehicle,UAV)影像匹配并行算法的负载均衡和特征数据传输问题,提出一种顾及空间邻接关系的影像匹配并行算法,旨在不降低原有匹配精度的前提下,最大限度地提高影像匹配效率.在特征提取阶段,根据影像空间邻接关系完成各对应节点的初始任务划分,随后进行稍细粒度的二次划分,以确定最终的特征提取任务.在任务调度时,根据计算节点的状态优先分配对应节点的任务.对应节点的任务全部分配完毕再分配其他节点对应的任务.在匹配阶段,首先根据特征提取任务划分匹配任务,然后根据特征提取任务的节点编号划定对应于每一节点的匹配任务单元,使用相同的任务调度方法完成整个测区的影像匹配.对一套有1 463幅UAV影像的典型数据进行实验,结果表明该算法不但能实现并行系统的负载平衡,还能减小特征数据的传输量,从而显著提高影像匹配效率.
[Abstract]:In order to solve the problem of load balancing and feature data transmission in UAV (unmaned aerial vehicle UAV image matching parallel algorithm, a parallel image matching algorithm considering spatial adjacency is proposed, which aims at not reducing the original matching accuracy. To maximize the efficiency of image matching. In the phase of feature extraction, the initial task partition of each corresponding node is completed according to the image spatial adjacency relationship, and the final feature extraction task is determined by a slightly finer granularity quadratic partition. In task scheduling, the task of the corresponding node is assigned first according to the state of the calculated node. All the tasks of the corresponding nodes are assigned and then the corresponding tasks of the other nodes are allocated. In the matching stage, the matching task is divided according to the feature extraction task, and then the matching task unit corresponding to each node is determined according to the node number of the feature extraction task. The same task scheduling method is used to complete the image matching of the whole measuring area. A set of typical data with 1 463 UAV images is tested. The results show that the proposed algorithm can not only achieve load balance in parallel systems, but also reduce the amount of transmission of feature data, thus significantly improving the efficiency of image matching.
【作者单位】：西安科技大学测绘科学与技术学院;武汉大学测绘遥感信息工程国家重点实验室;
【基金】：国家重点研发计划基金(No.2016YFB0502200) 国家自然科学基金(No.41127901)资助
【分类号】：TP391.41

【相似文献】