基于AKAZE算法的大尺寸电路板图像拼接研究

Research on large⁃size circuit board image stitching based on AKAZE algorithm

针对电路板自动测试系统中相机畸变、镜头视野有限所造成的图像失真、模糊不全等问题,文中提出一种基于AKAZE特征检测算法的大尺寸电路板图像拼接方法。首先通过AKAZE特征检测算法分别提取多张局部重合的电路板图像的关键点并对其描述;然后采用随机抽样一致性算法对关键点进行筛选(达到亚像素级别),再经暴力匹配对电路板图像进行视角变换后,匹配各局部电路板图像的特征点;最后对已匹配的各电路板图像的重合部分按权重进行融合,从而拼接成完整的电路板图像。实验结果表明,经图像拼接后的大尺寸电路板图像与原图像相比,图像更富有细节,无畸变且图像高度完整。文中的图像拼接方法所产生的误差小于0.1 mm,拼接图像与原图像的重复率达98%以上,并且拼接后的大尺寸电路板图像更富有细节、更加清晰,说明该方法可提高电路板自动测试系统的准确率,并为后续的数据识别工作提供有效支持。

In allusion to the problems of image distortion and incompleteness caused by camera distortion and limited lens field of view in the circuit board automatic test system,a large⁃size circuit board image stitching method based on AKAZE feature detection algorithm is proposed.The key points of multiple locally overlapping circuit board images are extracted and described by means of the AKAZE feature detection algorithm,the key points are screened(to sub⁃pixel level)by means of the random sample consensus(RANSAC)algorithm,and the visual angle of the circuit board image is transformed by the brute⁃force matching to match the feature points of each local circuit board image.The overlapping parts of the matched circuit board images are fused according to their weights,so that the complete circuit board image is stitched.The experimental results show that,in comparison with the original image,and the large⁃size circuit board image after image stitching has more details,no distortion and high completion.The error caused by the image stitching method proposed in this paper is less than 0.1 mm,the repetition rate between the stitched image and the original image is more than 98%,and the stitched images of large⁃size circuit board are more detailed and clearer.It can be seen that this method can improve the accuracy of the circuit board automatic test system and provide effective support for the subsequent data identification work.