一种弹孔自动识别算法的研究  

An automatic recognition algorithm for bullet holes

作　　者：张旭光[1] 孙巍[2] 韩广良[1] 王延杰[1] 

机构地区：[1]中国科学院长春光学精密机械与物理研究所 [2]吉林大学测试科学实验中心,吉林长春130025

出　　处：《光学精密工程》2005年第6期747-753,共7页Optics and Precision Engineering

摘　　要：提出了在室外环境下从胸环靶图像中实时提取弹孔的算法。采用灰度投影法从靶位图中提取胸环靶有效区域,利用数学形态学突出弹孔特征,排除数字的干扰,并克服弹孔间的部分重叠,最后利用模糊理论准确地识别出弹孔。识别弹孔时利用直方图得到阈值,从而去除不可能为弹孔的点以减少计算。提出判断周围像素分布的方法解决区域内脱靶问题。结果表明,本算法可准确地提取胸环靶图像中的弹孔,对不同环境有着较高的适应能力。对于大小为640×480的图像,计算时间仅约150 ms,并且可以识别部分重叠的弹孔。An algorithm to extract the bullet holes from a chest bitmap out of doors real-time was proposed. A gray projection algorithm was used to extract the chest bitmap from the target bitmap. The mathematical morphology was used to stress the character of bullet holes and to overcome the disturbance of figures. Using mathematical morphology could resolve the overlapping of the bullet holes, then the fuzzy theory was used to extract the target holes accurately. For recognising the bullet holes, the histogram of the target bitmap was used to gain a threshold to exclude the pixels which could not become bullet holes to reduce the complication of computation and a strategy judging the distribution of the pixels around the target holes was used to exclude the invalidation bullet holes. Experiment results show that the algorithm can extract bullet holes accurately and adapt to various circumstances. If the size of the bitmap is 648 ×480, it only needs 150 ms to complete this algorithm. The new algorithm can also overcome the overlapping of the target holes.

关 键 词：靶位图 胸环靶图 数学形态学 隶属度 灰度投影 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]