基于膨胀运算的移动对象兴趣点检测方法

发布时间：2018-07-04 14:05

本文选题：信息处理技术 + 轨迹数据挖掘　；参考：《兵工学报》2017年10期

【摘要】：针对传统兴趣点检测算法在准确性和效率方面的不足,提出基于膨胀运算的移动对象兴趣点检测方法(DMDO)。通过矩阵二值化操作滤除停留点噪声,提高预测准确率,并用膨胀运算替代传统方法中的聚类算法提高算法效率。将DMDO在开放空间数据集AMSA和IMIS3Days上进行仿真实验,结果表明:DMDO相比基于密度的空间聚类算法,在数据集AMSA上准确率平均提高17.94%,算法效率提高6.63倍;在数据集IMIS3Days上准确率平均提高19.98%,算法效率提高9.13倍;相比以聚类点排序结果确定聚类结构算法,DMDO在数据集AMSA上准确率平均提高20.04%,算法效率提高14.61倍;在数据集IMIS3Days上准确率平均提高16.60%,算法效率提高42.19倍;DMDO相比传统方法均表现出较高的预测准确性、较低的时间开销,适用于解决大数据背景下的移动对象兴趣点检测问题。
[Abstract]:Aiming at the shortcomings of the traditional interest point detection algorithm in accuracy and efficiency, a moving object interest point detection method (DMDO) based on expansion operation is proposed. In order to improve the prediction accuracy, the binary matrix operation is used to filter the resident point noise, and the expansion operation is used to replace the traditional clustering algorithm to improve the efficiency of the algorithm. DMDO is simulated on open spatial data sets AMSA and IMIS3Days. The results show that compared with density-based spatial clustering algorithm, the accuracy of DMDO on AMSA is 17.94 and the efficiency of the algorithm is 6.63 times higher than that of density-based spatial clustering algorithm. The accuracy on the data set (IMIS3 days) was improved by an average of 19.98, the efficiency of the algorithm was increased 9.13 times, and the accuracy of the clustering structure algorithm was increased by 20.04 times and the efficiency of the algorithm was 14.61 times higher than that of the clustering structure algorithm on AMSA. On the data set (IMIS3 days), the accuracy is improved by 16.60, and the efficiency of the algorithm is improved by 42.19 times. Compared with the traditional methods, DMDO has higher prediction accuracy and lower time cost. It is suitable to solve the problem of moving object's point of interest detection under big data background.
【作者单位】：中国科学院电子学研究所;中国科学院空间信息处理与应用系统技术重点实验室;中国科学院大学电子电气与通信工程学院;
【分类号】：TP311.13

【相似文献】