摘要
目标形状因子和长、宽、高等几何特征,与目标类别具有很大的关系,而灰度直方图也是目标物理特性的直接表现。根据采用马尔可夫随机场理论分割后的声纳目标和阴影图像,利用其灰度直方图曲线进行目标的聚类分析,再利用几何特征进行类别识别,经实测的数据验证,取得了较好的效果。证明该方法可有效表征不同目标的物理特性,从而区分不同类别的目标,避免了目标绝对反向散射强度的复杂计算,增强了抗噪性。以搜寻目标的几何特征为输入参数,可迅速锁定要搜寻的目标,借助其它探测手段,实现目标的自动识别。
Object recognizing technology is introduced based on the histogram and geometric features in this paper. Geometric features include the length, the width, the height of object, etc. They have great relation with the class attribute of object. The physical character is also reflected by the histogram of object. After the object and its shadow have been segmented based on the MRF theory, their histogram curves are extracted from combined zones of the object and shadow. Then the correlative coefficients are calculated from histogram curves. The object can be clustered through the coefficients. And then, the sought object is can be recognized according to geometric features. Finally, the real data verify the algorithm. It shown that the algorithm is robust, can rapidly locate the object, and can automatic all recognize the object.
出处
《海洋通报》
CAS
CSCD
北大核心
2006年第5期64-69,共6页
Marine Science Bulletin
基金
国家自然科学基金资助项目编号:(40506023)
"基础地理信息与数字化技术"山东省重点开放实验室资助项目编号:(SD040212)
关键词
直方图
几何特征
声纳图像
目标识别
Histogram
Geometrical Feature
Sonar Image
Object Recognition
