Receiver’s Winding And Voice Coil Coating Defects Recognition Based On Machine Vision And Its Application

By converting audio electrical signals into sound signals, the receiver is widely used in mobile phone, telephone and other communication terminal equipments. In the manufacturing process, the defect of winding and voice coil coating is easily produced, which will reduce product quality and influence usage result. In this paper, we study the receiver’s winding and voice coil coating detection, which contains three key techniques and a set of application system:potential defect region segmentation, shape feature extraction and selection of winding and voice coil coating defect region, asynchronous classification and recognition of winding and voice coil coating and by which visual defects automatic detection system. Finally, the development of the system was tested and verificated.In the first chapter, the machine vision inspection system is introduced, and the research status of defect inspection technology based on machine vision is introduced. The characteristics and difficulties of surface detection, shape detection and impurity detection are analyzed. This paper describes the background and significance of this study, and gives the main research content and framework.The second chapter, by analyzing the regional characteristics of the receiver, fusion chromaticity histograms of local and natural feature of the receiver latent defect region segmentation method is proposed. The color image segmentation method that based on region growing is studied, by using correlations less components in HSV color space, the image is converted to HSV color space for segmentation, the obtained region of natural characteristics is studied. On the basis of this new method extraction method based on local color histogram receiver of the defect area, by drawing HSV component histogram of potential defects region, color spatial distribution and chromaticity information statistics is obtained. The receiver online segmentation that based on regional natural features is proposed. The local color histogram statistics is used for coarse segmentation on receiver region, and then the binary image connected characteristics and position information is combined for fine segmentation on receiver defects. At last, the potential defect region segmentation process of receiver is realized.The third chapter, based on the defect area finely segmentation, the shape features of receiver defects are extracted and selected. According to the classification of regional shape features, it is divided into internal shape and external shape features, and the method of obtaining the characteristics of each shape is given. On the basis of analysis of existing feature descriptor, aiming at receiver block defect area and linear defect area, the defect area is optimized by mathematical morphology, then the binary area feature is extracted, the shape features is selected for subsequent classification by criterion of features.The fourth chapter, the component of typical pattern recognition system is studied, the classification system of asynchronous receiver is established. By analysing the detection requirements of defect area, the classification method of asynchronous receiver winding and voice coil coating defects is proposed. The linear pattern recognition method is used to rapidly identify whether there is image defects, analysing defect area characteristics, different regions takes advantage of different characteristics to build BP neural network, the constructed neural network can accurately identify the defects. By analysing the results of the detection and assessmenting the performance, the effectiveness of the method are validated.The fifth chapter introduces the automatic visual detection system of receiver defects and its application verification. The key factors of hardware and software design are described in this paper. The fusion chromaticity histograms of local and natural features of the region segmentation method, the shape feature extraction and selction of defect area and defect asynchronous classification recognition are applied in the detection of defects of receiver, a automatic visual detection system of receiver defects is developped. The receiver detection application prototype system is applied to validate the method and technology.In the last chapter, the further research is put forward after summarizing all the content and achievements of the dissertation.