基于形状描述直方图的声呐图像目标识别算法  

Mine target recognition algorithm of sonar image

作　　者：田晓东[1] 刘忠[1] 周德超[1] 

机构地区：[1]海军工程大学电子工程学院,湖北武汉430033

出　　处：《系统工程与电子技术》2007年第7期1049-1052,1212,共5页Systems Engineering and Electronics

摘　　要：针对水下声呐图像目标检测与识别问题,在图像规格化条件下,定义了几种不同的形状描述子,并统计计算描述子直方图,定义了几种不同的直方图匹配测度函数用以计算直方图间的相似性。分析了算法中关键参数的选取问题。仿真结果表明,基于该算法的识别率可达到98%以上,具有较强的不变性和鲁棒性。通过进一步信息融合算法可快速准确地实现水下声呐图像目标识别,具有较强的应用价值。In the target detection and recognition of underwater sonar image, the recogmtlon of mine targets is one of key technologies. With analysis and calculation, based on image normalization, several shape descriptors are defined, then their statistical histogram are calculated. Several different measurement criterions are presented to calculate the similarity between two histograms. Important parameters should be selected carefully in this algorithm, therefore this crucial problem is especially analyzed. Simulation result indicates that the algorithm presented has characteristics such as precision and robustness, the correct ratio can be above 98%. Thus it has great value of application.

关 键 词：目标识别 形状描述子 直方图 声呐图像 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]