Research On On-line Defect Detection Algorithm For Color Printing Based On Machine Vision

With increasing of the requirement of printing’s quality and printing speed,the traditional manual or off-line detection methods can not meet the real-time controlling and the needs for high-quality printed materials.High-performance automatic printing detection system is an effective way which can reduce scraps and improve print quality and efficiency.With the study of classic detection algorithm,the automatic printing defect detection system based on artificial neural network is designed.Aiming at the on-line detection of defect classification,the module is proposed based on genetic algorithm optimized neural network strategy.In this paper,overall scheme of on-line defect detection system for color printing solutions based on artificial neural networks is designed firstly.Each module’s function of hardware and software is introduced in detail,and development and advantage of on-line defect detection system are discussed.Secondly,the defect extraction module based on dynamic threshold is studied.The drawbacks of traditional extraction algorithms are discussed,and the defect extraction based on improved dynamic threshold algorithm is put forward.Through the performance of the algorithm of experiments is analyzed and validated,experimental results show that the algorithm reaches 98%in accuracy,and efficiency is 26%higher than traditional threshold algorithm.The algorithm can fast detect all types of defect location and size,meeting requirements of the print comprehensive online quality.Thirdly,by analyzing advantages and disadvantages of traditional defect classification methods,the new defect classification method based on the genetic algorithm optimization BP neural network is proposed.The shape classification and color classification of printing defect are both achieved.Through experiments,the neural network model is trained and tested,and experimental results show that the network model can meet the application requirements.Finally,the processing software of color printed matter defects inspection is prepared by VisualC++6.0,based on the library OpenCV.The image preprocessing is realized,and precise defect extraction and accurate classification are achieved by using the improved algorithm.