基于改进直方图不变矩的声呐图像识别方法  

Recognition for sonar image based on improved histogram invariant moments

作　　者：赵春晖[1] 李誉斐[1] 

机构地区：[1]哈尔滨工程大学信息与通信工程学院,黑龙江哈尔滨150001

出　　处：《应用科技》2007年第11期19-24,共6页Applied Science and Technology

基　　金：国家自然科学基金资助项目(60672034);高等学校博士学科点基金资助项目(20060217021);黑龙江省自然科学基金重点资助项目(ZJG0606-01)

摘　　要：针对声呐图像受噪声干扰大,目标边界较模糊的特点,并结合传统直方图不变矩识别图像快速的优点,以及其不考虑灰度空间信息而易造成误识的缺点,将目标的形状因子加入不变矩特征向量中,提出了一种对亮度和对比度变化不敏感,具有旋转和尺度不变性的改进的直方图不变矩.实验仿真结果表明,该方法能够快速、有效地识别水下目标.Considering the blurred boundary on the sonar image of target due to the heavy noise, this paper put forward an improved histogram invariant moments for the recognition of sonar image in conjunction with the shape factor from the traditional histogram invariant moments. This method is not sensitive to luminance and contrast. Mo- reover, the histogram invariant moments in this method feature invariant rotation and scale, not only preserving the recognition speed of the original method but avoiding the probability of wrong recognition due to the space informa- tion loss. Experimental results show that the improved method can recognize the underwater target quickly and effectively.

关 键 词：声呐图像 识别 特征提取 直方图 不变矩 形状因子 

分 类 号：TP919.8[自动化与计算机技术]