Study On Detection Method Of Solder Joint Based On Principal Component Analysis And Support Vector Machine

With the demand of printed circuit board to high-precision and multi-layer, solder joint as the bridge of connection between each electric appliance components and circuit board plays an important role in the production of industry. Aiming at the influence on the quality of solder joint caused by the instability of welding equipment and the uncertainty of operating environment, solder joint defects detection based on visual information has become a research hot spot in machine vision.Firstly, based on the wavelet theory an enhanced algorithm of wavelet theory is introduced in this thesis, and this algorithm enhances some effective wavelet coefficients in the process of wavelet decomposition. The wavelet enhancement algorithm effectively reduces the computational complexity of wavelet theory and refines wavelet coefficients. Then the principal component analysis combined with wavelet coefficients is proposed in this thesis to extract features of solder joint image. The extracted 50 dimension of wavelet coefficients characteristic vectors are reduced by using principal component analysis. And the wavelet characteristic vectors of dimension reduction are combined with the shape feature of image and histogram feature, thus the effective characteristic values of solder joint are extracted.In addition, considering the factors of reducing the solder joint training samples and nonlinear characteristic values of solder joint, the decision-tree combined with improved support vector machine is also proposed to identify the types of solder joint defects. The experimental results show that the extracted characteristic values of solder joint has better discrimination and compactness. And the proposed classification method effectively improves the recognition rate of the solder joint defects detection and the efficiency of the algorithm.Finally, the friendly human machine interface of solder joint defects detection is designed in the environment of MATLAB GUI. The operator can set inbuilt parameters according to industry needs. In order to simplify user operation, one-touch-recognition processing module and effective system platform are provided, the operator can achieve real-time control in the process of solder joint defects detection.