The invention provides a
product design closed-
loop optimization method combined with defect detection feedback, which comprises the following steps of: data collection: collecting a
product image and marking an actual defect, and constructing a defect detection
data set; defect detection model training: training the first
deep learning model to obtain a defect detection model;
data analysis: analyzing a detection result by using a defect detection model, identifying high-frequency
false detection and missing detection, and associating to design features causing problems; training an aided design model: according to an analysis result, pointedly labeling easily-confused design features, constructing an aided
design data set, and training a second
deep learning model to obtain the aided design model; and performing design iteration: modifying or optimizing easily-confused features in a design drawing through
design software by utilizing an aided design model or an analysis result. According to the invention, by establishing closed-loop feedback, characteristics which are easy to cause detection errors are actively avoided in a
design stage, so that the detection precision is improved, the cost is reduced, the period is shortened, and the product quality is improved.