基于变换矩阵的BGA X-ray图像倾斜识别

针对基于X-Ray的焊球阵列封装(BGA)焊接缺陷检测中，现有的图像处理算法无法有效处理PCB意外倾斜的问题，提出了一种自动判断BGA焊点X射线图像倾斜程度并进行校正的方法。首先根据BGA焊点阵列排布的特性恢复出原始图像中的结构信息，然后使用广义逆矩阵求解出理想正视图像与实际倾斜图像之间的变换矩阵，最后对倾斜图像进行逆变换以得到理想正视图像，并使用变换矩阵估计出PCB的倾斜程度。实验针对正视与倾斜的BGA焊点的X射线图像，运用所提方法，准确判别出图像的倾斜程度，并进行了校正。此方法既可以作为一种图像质量评价算法，在BGA X-Ray缺陷检测中判断图像是否为需要的正视图像；也可以作为一种图像自动校正算法，提升基于X射线的自动检测系统的适应性。

A recognition and correction method of declining BGA X-ray image based on transformation matrix

During the quality inspection of BGA solder balls based on X-ray images, existing image processing algorithm can not effectively deal with the images of unexpectedly inclined PCB. A method which could automatically judge the inclination of X-ray images of BGA solder balls and correct them was proposed. The feature of BGA solder balls array was used in the method to rebuild the structural information firstly. Then the transformation matrix between ideal front view image and real inclined image would be calculated by Moore-Penrose generalized inverse matrix. Finally, the inverse matrix of transformation matrix was used to get the ideal front view image. At the same time, the angle of PCB could be estimated by using the transformation matrix. When used in the inclined X-ray images of BGA solder balls, the method could judge the inclination effectively and correct it to get the front view image. The method can be used as an image quality evaluation algorithm to judge whether the image is a desired image in industrial detection, as well as be used as an automatic image correction algorithm to improve the adaptability of the automatic detection system based on X-ray.