基于聚类与SIFT算法的自动地形成图方法研究  

Research on Automatic Graph Forming Method Based on Clustering and SIFT Algorithm

作　　者：范仲雪 张东升[1,3] 曹金亮 FAN Zhongxue;ZHANG Dongsheng;CAO Jinliang(Marine Acoustics Technology Center,Institute of Acoustics,Chinese Academy of Sciences,100190,China;University of Chinese Academy of Sciences,100049,China;Beijing Marine Acoustic Equipment Engineering Technology Research Center,100190,China)

机构地区：[1]中国科学院声学研究所海洋声学技术中心,北京100190 [2]中国科学院大学,北京100049 [3]北京市海洋声学装备工程技术研究中心,北京100190

出　　处：《网络新媒体技术》2022年第1期27-34,共8页Network New Media Technology

摘　　要：针对智能水下机器人由于长时间工作,在进行地形图构建时,成图精度变差的问题,提出了一种基于聚类和尺度不变特征转换算法的,可以对相邻测线的地形图进行自动成图修正的新方法。利用尺度不变特征转换算法和K-means算法进行特征区域的提取与配对,再利用卡尔曼滤波算法进行目标跟踪,即可得到较为准确的坐标数据,从而提高地形图精度。试验证明该方法能够较为准确地进行特征点的选择以及配对,并进行目标跟踪,将误差降低56.15%。实现了相邻测线的自动修正和地形成图,具有可行性。In order to solve the problem that after long time working the navigation accuracy of autonomous underwater vehicle will affected by the cumulative error, this problem leads to poor accuracy of the geomorphological map. Therefore, this article put forward a new method use clustering and scale-invariant feature transform to automatic correct geomorphological map of neighboring survey line. The proposed method mainly includes, using SIFT and K-means algorithm can extract feature regions, then using Kalman filter can track the target to get more accurate coordinate date and improve the map accuracy. Under the method above, experiments show that the method can select and match feature more accurately and perform target tracking, and made the error reduce 56.15%. This method realized the automatic correction and mosaic the map of neighboring survey line, it can effectively improves the accuracy of the mosaic map.

关 键 词：水下地形成图 目标跟踪 聚类 尺度不变特征转换算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP311.13[自动化与计算机技术—计算机科学与技术]