Research On The Recognition Method Of Solder Joints Based On Wavelet Transform

With the development of printed circuit board assembly technology to higher density and‘zero defect’, the market requires the automatic optical inspection system to be higher accuracyand intelligence. So that solder joints defect detection technology has been a research hot spotin this field. The key of solder joints defect detection technology based on image processing isselecting the appropriate method of image edge feature extraction and recognition. Extractingeffective solder joints feature information is the premise condition of the defect detection.This paper proposes an improved wavelet enhancement algorithm based on the originalwavelet theory. This algorithm adopts strengthen function of piecewise linear, only bystrengthening a certain part of the wavelet coefficient, and the strengthen function is selectedadaptively with the scale’s difference. The experimental results show that the improved waveletenhancement algorithm reduces the amount of calculation, reflects the details of the waveletcoefficients better, improves the recognition rate of solder joints defect greatly, and obtainedsthe satisfactory results.In addition, this paper identify the extracted feature vector of the solder joints by BP neuralnetwork, RBF neural network, one to more SVM classification method and binary tree SVMmethod. Comparing their test results, we can find that when the sample size is small, thesupport vector machine method has the higher recognition rate; When the sample size bigenough, the neural network method has the slightly higher recognition rate than the supportvector machine method.Finally, this paper designs a detection system interface of solder joints defect, which isbased on a series of software development of solder joints defect detection. By it, the user couldunderstand the process of solder joints feature extraction and defect recognition more intuitive,and set the parameters which their need.