基于PCA和粒子群优化算法的焊点缺陷识别

Solder Joint Defect Recognition Based on PCA and Particle Swarm Optimization Algorithm

摘要:

针对生产线上的表面贴装技术(SMT）焊点图像的特点，提出了一种基于PCA和粒子群算法-误差反向传播（PSO-BP）神经网络的焊点缺陷识别方法。首先使用图像处理技术和CCD传感器对PCB焊点图像进行预处理，采用中值滤波、灰度图像增强、全局阈值法等方法，有效抑制噪声干扰并提高了图像对比度，提取出较好的图像特征。然后运用主成分分析法提取包含焊点86.6%特征信息的5个主成分，并输入到经粒子群算法改进后的BP神经网络。通过具体的实验分析，结果表明改进的BP神经网络具有较好的识别分类效果，能够对正常、多锡、少锡、漏焊四种不同类型的焊点进行识别，准确率达93.22%，算法可靠，在实际生产中能够有效的提高检测效率。

Abstract:

Aiming at the characteristics of surface mount technology (SMT) solder joint image on the production line, a solder joint defect identification method based on PCA and particle swarm optimization algorithm-error backpropagation (PSO-BP) neural network is proposed. Firstly, image processing technology and CCD sensor are used to preprocess PCB solder joint image. Median filtering, gray image enhancement and global threshold method are used to effectively suppress noise interference and improve image contrast, and extract better image features. Then, the principal components analysis method is used to extract the five principal components including the 86.6% characteristic information of the solder joints, and input into the BP neural network improved by the particle swarm optimization algorithm. Through specific experimental analysis, the results show that the improved BP neural network has better recognition and classification effect, and can identify four different types of solder joints of normal, multi-tin, less tin and miss welding, with an accuracy rate of 93.22%. Reliable, it can effectively improve the detection efficiency in actual production.