Abstract: This paper addresses the problem of low efficiency and reliance on manual inspection for surface quality assessment after fine cutting of glass cover plates, an automated defect detection system for glass cover plates based on a feature cascade network was designed. The system is composed of a mechanical structure, a vision system, and a control module, and its workflow includes barcode information reading, cutting accuracy measurement, front-and-back defect detection, and visual positioning for material handling. By improving the mechanical structure design and vision processing algorithms, and combining a cascaded BP neural network with a CNN-based coarse-to-fine classification approach, the system achieves standardized comparison and precise classification of multiple defect types. The results show that the system can effectively enhance the accuracy and efficiency of defect detection in glass cover plates, providing reliable technical support for quality control of smartphone and similar device glass screens.