Technology And Research Of Button Defects Detection Based On Machine Vision

In the production process of buttons,buttons are prone to various defects due to uncontrollable factors such as mechanical failure and mold damage.At present,the detection method of the button is usually manual visual inspection,and the method is inefficient and limited in accuracy.Aiming at this problem,this paper is based on the relevant theoretical knowledge of machine vision,and considering the characteristics of different types of buttons,this paper proposes a method based on machine vision for the detection of plastic buttons and metal buttons.This paper builds a hardware platform for button detection and mainly studies three parts: first,the segmentation of the region of interest of the button image;second,the detection method of the plastic button;and the detection method of the metal button.The main contributions and innovations of the thesis are as follow:1.A region image segmentation algorithm based on morphological processing and Graham algorithm is proposed.The algorithm filters the collected button images.According to the different characteristics of the plastic buttons and the metal buttons,the morphological processing is used to obtain the area where the button is located in the image,and the edge of the button is extracted.Finally,the Graham algorithm is used to obtain the coordinates of the minimum circumscribed matrix of the button,and the coordinate value is mapped to the position where the original image is located,and the interest region of the button image is segmented.The algorithm lays a foundation for button defects detection,removes irrelevant interference information,and improves the efficiency of defects detection.2.A plastic button defects detection algorithm based on dynamic threshold and adaptive Gaussian clustering is proposed.The technology is divided into two modules according to the different types of defects: firstly,For the characteristics of button shape,this paper uses dynamic threshold segmentation and connected area counting to detect the inner hole defects of plastic buttons,and uses edge extraction and roundness analysis to detect the contour of plastic buttons.Secondly,according to the color difference between the defects region and the non-defects region,this paper uses the adaptive Gaussian mixture model to cluster the color of the button surface,and judges whether the plastic button has surface flaw based on the mean matrix obtained by clustering.It is found through experiments that the algorithm has accuracy and superiority for the detection of plastic button defects.3.A metal button defects detection algorithm based on categorical sparse representation classification of extreme learning machine is proposed.The algorithm transforms the detection problem into a pattern classification problem.In the preprocessing stage,the Exemplar-based algorithm is used to de-reflect the button image.The cascade classifier proposed in this paper is divided into two stages: First,because the extreme learning machine has the advantages of fast learning ability and strong generalization ability,the limit learning machine is used to classify the buttons to detect the defective products and the qualified products;In order to further improve the accuracy of the button classification,when the output probability of the single hidden layer feedforward neural network is lower than the set threshold,the classification algorithm is used to reclassify the button,and the dictionary is obtained by querying the sparse representation.Final classification results.It can be seen from the test results that the algorithm in this paper can detect the button products well and overcome the limitations that the single model cannot balance between computational complexity and classification accuracy.