Research On Carrier Tape Defect Detection Based On Machine Learning

As ribbon industrial products,carrier tapes are generally used for carrying electronic components.Holes and positioning holes are equally spaced in the length direction of carrier tapes,between which the distance is same,together with the shape.The surface quality of the carrier tape directly affects the mounting equipment catching the electronic components.Therefore,the defect detection technology of the carrier tape has drawn more and more attention.Although there are some devices been used to detect the carrier tapes,currently there are few related literatures on the defect detection of the carrier tape,and there has been no research on the carrier tape defects classification.In this paper,the image denoising technology is introduced firstly,which is using the combination of median filtering and wavelet transform denoising to denoise the noise of the carrier image.After that,it will comes to a duplex vision system based on machine vision.The system is consist of photoelectric sensor,amplifier,industrial camera,lens,industrial computer,monitor,light source,main control panel,motor,light alarm device and image analysis software,etc.This paper presents a hardware design scheme and a mechanical structure scheme,and designs an image processing algorithm suitable for carrier tape detection.The algorithm takes the standard qualified carrier tape sample images as the training set,and obtains the featured information of the image of the detection area from these qualified carrier tapes' sample images.Then the featured information serves as a comparison standard for on-line inspection.After researching on the two-position tape carrier system,useful images of defected carrier tape will be obtained.On this basis,a method obtaining texture features and morphological features is used to extract the features of the flawed images.Since the original feature extracted directly has irrelevant features and redundant features,the original feature is filtered by a combination of ReliefF algorithm and clustering algorithm.Finally,the supported vector machine is used totrain the extracted features,and the improved support vector machine classifier is used to classify the carrier-borne defects.