| Summary: | Glass wafer (GW) is used in a variety of integrated circuit (IC) packaging applications and as substrates to provide better performance and cost-effectiveness. Glass wafer (GW) protects the IC from impact and corrosion while maintaining the contract pins and leads that connect it to the external circuit. In the process of technology or production, this kind of structure is continuously working for a long time. Due to the inherent defects such as bubbles generation, starvation and the structure is often suffered from acid, alkali, moisture, vibration and other factors, which makes its internal structure gradually form corrosion stains. These defects have posed a serious threat to the quality and performance of equipment. Based on these disadvantages, this paper analyzes the defect detection principle of Glass wafer, then designed a method of determine the defect region. The edge signal processing method of visual image defects is studied, the edge detection and defect feature extraction model is established, and the principle of defining strong and weak edges is clarified. A defect feature classifier based on multi-layer perceptron (MLP) is created, and a segmentation algorithm of the classifier is implemented. Finally, a multi-channel image detection experimental platform is built to verify the typical unit structure. The experimental results show that the rate of defective features recognition is high, the detection rate is fast, and it has practical application value in engineering. The research of this recognition method has positive theoretical significance for accurately evaluating the overall reliability of GW structure and ensuring the safe operation of equipment.
|
|---|