Electrolytic capacitor identification and polarity detection based on deep learning
Aiming at the problem that electrolytic capacitors on printed circuit board(PCB)in engineering are difficult to identify and their polarity is difficult to detect,an electrolytic capacitor identification and polarity detection method based on deep learning is studied.For the capacitance detection problem,an improved algorithm based on YOLOv5 is proposed.This algorithm integrates the Swin Transformer module on the backbone of YOLOv5 to improve the feature extraction capability of the model,and integrates the bi-directional feature pyramid network(BiFPN)module on the neck to improve the feature fusion capability of the network.For capacitor polarity detection,a method of semantic segmentation combined with morphological processing is proposed.This method can segment the base circle area and polar area of the capacitor and then effectively detect the polar direction of the electrolytic capacitor.Experimental results show that the capacitor identification precision reaches96.9%,the capacitor segmentation precision reaches93.9%,and the polarity direction detection accuracy reaches99.1%.Compared with the current detecting method of the polarity of electrolytic capacitors,the proposed method has better robustness and meets the detection requirements.
deep learningelectrolytic capacitorpolarity detectiontarget detectionsemantic segmentationelectronic component
万方数据
维普
