Research On Solder Joint Detection Based On Haar Feature And AdaBoost Algorithm

With the development of modern science and technology, electronic products increasingly become digital, and the printed circuit board also become component-intensive and miniaturized. But the traditional detection technology in effect and speed can not meet the needs of the new SMT technology. Reliable and fast automatic detection of the printed circuit board has become an important technology to improve the Electronics manufacturing automation level and the product quality. The automatic detection based on machine vision technology more and more get the attention of people. The detection technology utilizes the computer technology, high-speed image processing and recognition technology, automatic control technology, precision machinery and optical technologies, it is a combination of a variety of high-tech products, with automation, high speed and high resolution detection ability. It greatly reduces the labor intensity, improves the objectivity and accuracy of criterion.In the printed circuit board detection, the solder joint detection is one of the important aspects. The position of the solder joint, and the presence of defects, can directly affect the printed circuit board to work properly, So, study in solder joint detection is currently a hot research topic. In the study of its detection, the method based on machine vision in a dominant position significantly, especially machine learning algorithms.Machine learning with its powerful generalization ability, can achieve very high accuracy in prediction, and the algorithm model is established in this paper, the process of learning can adaptively optimize, even with the increase of sample, can be secondary to learn in the process of detection. Applying machine learning into the printed circuit board solder joint detection, can not only solve the artificial detection speed, low productivity, but also can reduce labor cost, improve the detection accuracy.The author's major contribution focusing on the solder joint inspection are outlined as follows:1. Process of the printed circuit board detection technology development is summarized, the theoretical research to Template matching algorithm, support vector machines and BP neural network algorithms in solder joint inspection is done.2. Detailed study of the Haar feature is done. more about the Adaboost classifier. the other two similar algorithms are analyzed, based on HOG feature and LBP feature. In Visual Studio 2013 development platform, with Open CV rich library of codes, the code is written. Adaboost classifier algorithm based on Haar feature is applied to the solder joint inspection. the samples to generate a classifier training are used, and the resulting classification is applied to the printed circuit board to the PCB solder joint inspection process, Adaboost algorithm based on Haar feature is Determined as a kind of good effect of solder joint detection algorithm.3. By focusing on the phenomenon of false detected solder joints, adding the color feature into Haar feature is been proposed. the optimal HSV model is selected as a color space, and then through the software implementation, color features is added to the Haar features in the sample training successfully. It is proved to be an efficient algorithm.