Research On Bubble Detection Technology Of BGA Solder Ball

With the rapid development of chip integration technology,BGA packaging technology is more and more popular,and the packaging environment is more and more complex,which leads to more and more complex background interference of X-ray image generated by X-ray flaw detection.As a result,BGA bubble detection based on traditional algorithm is gradually not suitable for the current environment.Therefore,in this paper,a series of researches have been carried out to solve the problems of complex background interference,low contrast of solder ball bubbles and false segmentation of edge bubbles in X-ray flaw detection of BGA solder ball.First of all,aiming at the problem that BGA ray image detection is easy to be interfered by complex background factors such as wire coil and chip,a method of extracting solder ball based on full convolution network is proposed.In this method,BGA tag data set is firstly made according to the target solder ball region;then a full convolution network is built,through which training parameters are adjusted on the network,the final appropriate network model is obtained;finally,the solder ball region is extracted from the image to be measured.Experimental results show that this method can remove the complex background factors completely,thus effectively extract the area of the target solder ball,and lay the foundation for the fine segmentation of post-processing.Then,in order to solve the problem that the edge bubble pixel level is close to the background pixel level,which makes the bubble segmentation of BGA solder ball inaccurate,a method based on K-means clustering is proposed.In this method,firstly,K-means clustering algorithm is used to gather solder balls,bubbles and background into different clusters;secondly,mathematical morphology and region growing algorithm are used to optimize the segmentation image;finally,BGA bubble segmentation image is obtained.Experimental results show that the method can completely separate the inner and edge bubbles of BGAsolder ball.Finally,the algorithm of defect identification is studied.Defect calculation and marking are carried out on the segmentation image of solder ball,and the qualified condition of solder ball in BGA image is counted.The results are compared with the traditional BGA segmentation algorithm.Experiments show that the method proposed in this paper has better adaptability and accuracy.