Research On Surface Crack Detection Method Of Tantalum Capacitor Terminal Electrode

The surface crack defect of the tantalum capacitor end electrode seriously affects the service life of capacitors.It is of great engineering significance to find the crack defect in time and reduce the defective rate in the manufacturing process,which is an important link to ensure the ex-factory quality of tantalum capacitors.In this paper,due to the problems of low resolution and high noise on the tantalum capacitor electrode surface due to the constraints of the actual detection environment,a selective gradient-guided image high fidelity super-resolution reconstruction preprocessing method is studied.In view of the problems that are difficult to detect due to the characteristics of not obvious cracks and small area proportion,a weak feature micro-crack identification method based on crack contour guided feature enhancement was studied to achieve accurate detection of surface cracks on tantalum capacitor electrode.Specific research work is as follows:Aiming at the problem of low resolution and high noise of the end electrode surface image,the selective gradient guided image reconstruction preprocessing method for crack under the condition of considering noise was studied to achieve the acquisition of high fidelity and high-resolution image to be detected.An image acquisition algorithm based on the shape features of the tantalum capacitor electrode surface was designed to accurately obtain the image of the region to be detected from the electrode surface image.An interpretable superresolution network model of the gradient-guided image is proposed to simulate the attention mechanism of the human visual system to approximate the effect of pixel-level divide-and-conquer high-fidelity image reconstruction.A pixel-level selection module for microcracks is designed and embedded into the image reconstruction model,and a high-fidelity reconstruction network model for selective cracks is constructed.The experimental results show that the proposed preprocessing method can accurately obtain the image to be detected,the reconstructed image peak signal-to-noise ratio is 44.71 d B,the structural similarity is 0.9771,and the detection time is about 8.19 ms.Moreover,the pixel-level selection module of micro-crack can reduce the influence of interference noise to a certain extent.Aiming at the problems that the crack is not obvious and difficult to extract and the feature is easy to be lost due to the small area ratio,the weak feature microcrack identification method based on contour-guided feature enhancement is studied to realize the high-precision detection of the surface crack of the opposite electrode.A contourguided pixel-level feature enhancement module was designed to extract,retain and strengthen crack features effectively in the same resolution feature layer by using information such as microcrack profiles.A multi-scale feature fusion module under the stepped-attention mechanism is proposed,and the multi-scale features in the feature enhancement module are extracted and fused under the stepped-attention mechanism.A microcrack identification and classification module was designed,and a microcrack identification model with a weak feature on the end electrode surface was constructed.The experimental results show that the accuracy of the method proposed in this paper is99.41%,the accuracy is 99.26%,the recall rate is 99.51%,the F1 fraction is 99.38%,and the area under the subject’s working characteristic curve is 99.95%,which meets the requirement of high precision crack detection.On the background of tantalum capacitor products provided by actual enterprises,a micro-service system for visual detection of tantalum capacitor end electrode surface cracks is built on the basis of the above methods,and a highly integrated image preprocessing method and weak feature micro-crack identification method is realized.The tantalum capacitor end electrode surface image data set is taken as a case for verification.The research results of this paper provide an important theoretical method for the automatic visual detection of surface cracks of tantalum capacitor end electrodes and have important practical engineering value for improving the factory quality of tantalum capacitors.