复杂背景图像中军用靶子识别算法研究  

Research of target recognition algorithm from image with complex background

作　　者：尚春红[1] 赵明昌[2] 

机构地区：[1]中国科学院国家科学图书馆,北京100190 [2]中国科学院自动化研究所,北京100190

出　　处：《计算机应用》2008年第5期1257-1260,共4页journal of Computer Applications

摘　　要：提出了一种新的靶子图像识别算法,专门针对野外实弹射击采集的、具有复杂背景的图像。首先利用颜色特征,通过RGB空间转换到HSI空间、S通道阈值分割、融合I通道信息、第二次阈值分割、形态学后处理等步骤,进行一次粗分割;然后利用区域特征,提出了一种基于AdaBoost学习算法的靶子分类器设计方法,可以较好地将靶子区域同其他杂质区域分开,得到最终识别结果。This paper proposed a new algorithm of recognizing military target from image captured by outdoor firing training system with very complex background. Firstly the color feature of target was utilized to do a coarse segmentation. The steps include converting from RGB color space to HIS color space, thresholding in S channel, combining I channel information, thresholding in combined image and postprocessing using mathematical morphology. Then a classifier based on AdaBoost learning algorithm was designed to separate target region from others and get the final recognition result.

关 键 词：彩色图像分割 不变矩 AdaBoost学习算法 分类器 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]