复杂背景图像中军用靶子识别算法研究

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。

Research of target recognition algorithm from image with complex background

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.