Study On The Defect Detection And Classification Of Surface Mounting Components Based On Machine Vision

With the rapid development and widespread use of modern electronics industry,the demand for chips continues to rise.The chip is prone to surface defects such as scratches and holes during the production process,which will directly affect its working performance.In order to strictly control the fraction defective of chip,the defect detection technology is particularly critical.Traditional detection methods are not only having low detection efficiency,but also affected by human factors,making it difficult to meet actual production requirement of modern industry.The machine vision technology has the advantages of high recognition accuracy,fast efficiency and non-contact.Applying this technology to chip surface defect detection can effectively solve the problems of traditional detection methods.Therefore,the researches on surface defect detection and classification technology based on machine vision are of great theoretical significance and application value.Firstly,an image acquisition system for m achine vision is designed according to the defect characteristics of the detected chip,and the selection basis of each hardware part is studied.Also the workflow of the software image processing algorithm is analyzed.Secondly,the image is preprocessed and then is segmented and extracted-Respectively using iterative method,maximum between-cluster variance algorithm and two-dimensional maximum entropy algorithm to threshold the image.The experimental results show that the two-dimensional maximum entropy algorithm can effectively segment the target defect area compared with the former two segmentation algorithms.However,due to the large amount of computation and long time-consuming problem,this paper proposes an improved genetic algorithm to optimize it quickly.The experimental results show that this method can accelerate the convergence speed of the algorithm,effectively shorten the calculation time and obtain the segmentation threshold which is better than the two-dimensional maximum entropy algorithm.Nevertheless,there are still some noise points in the segmented image.The morphology and connected area method are used to remove the noise and extract the defect part.On this basis,the defect samples are extracted from the geometric and texture features,and the original features are reduced by PCA.Besides,the improved genetic algorithm is used to optimize the parameters in the support vector machine,and the dimensionality-reduced feature vector is used as the input of the support vector machine classifier to train the samples.Finally,through multiple defect classification test experiments,the average recognition accuracy rate can reach 93.88%.The results show that the detection method in this paper can effectively classify the chip surface defects and has certain application value.