CAJ | 학술논문

针对当前采用目测识别方法进行炮膛疵病检测时存在的识别速度慢、精度低、人为影响因素大等缺点，提出了一种基于拼接算法的炮膛疵病检测方法。该方法采用Harris角点检测算法进行特征点提取，采用归一化互相关法进行特征点初匹配，采用马氏距离提纯算法消除误匹配，采用改进的8参数透视变换优化估计算法和基于边界保持的函数加权平滑融合算法进行炮膛图像融合，并选取图像面积、短长径之比为特征参数完成疵病的自动识别。实验结果表明：该算法对炮膛图像能较好地进行拼接，而且数值化的疵病识别方法相比传统目测识别方法也更加快速、准确，可为部队和相关工程人员进行身管检测提供有益参考。
침대당전채용목측식별방법진행포당자병검측시존재적식별속도만、정도저、인위영향인소대등결점，제출료일충기우병접산법적포당자병검측방법。해방법채용Harris각점검측산법진행특정점제취，채용귀일화호상관법진행특정점초필배，채용마씨거리제순산법소제오필배，채용개진적8삼수투시변환우화고계산법화기우변계보지적함수가권평활융합산법진행포당도상융합，병선취도상면적、단장경지비위특정삼수완성자병적자동식별。실험결과표명：해산법대포당도상능교호지진행병접，이차수치화적자병식별방법상비전통목측식별방법야경가쾌속、준학，가위부대화상관공정인원진행신관검측제공유익삼고。
In allusion to the shortcomings of slow detection speed, low accuracy and great artificial factors existed in the current artillery bore flaw recognition by visual detection, a method of artillery bore flaw recognition based on jointing algorithm is proposed. The method extracts feature points by using Harris corner detection algorithm, matches feature point by using the normalized cross correlation method, elimi- nates false match by using Ma distance purification algorithm, fuses the bore image by using improved 8- parameter perspective optimization estimation algorithm and boundary weighted smoothing fusion algo- rithm, and takes image area and the ratio of short and long diameters as the characteristic parameters to complete automatic identification of defects. The experimental results show that the algorithm is good for bore image jointing, and the numerical flaw image recognition is more rapid and precise than the tradi- tional visual detection, which can provide a useful reference for army and engineering staff.