融合计算机视觉的模具裂纹自动检测与精细分割算法

打开文本图片集

中图分类号：TG76 文献标志码：A

Computer vision-based automatic detection and fine segmentation algorithm for mold cracks

YUAN Hui

（Shaanxi Polytechnic Institute，Xianyang 712Ooo，Shaanxi，China）

Abstract： Mold crack detection is an important link to ensure product quality and production safety in industrial manufacturing. In this article， an automatic detection and fine segmentation algorithm of mold cracks combined with computer vision is presented， aiming at realizing efficient and accurate crack identification and location. Firstly，the crack image data set is expanded by using deep convolution generation countermeasure network （DCGAN） to solve the problem of insufficient samples. Then，an improved convolutional neural network （CNN） is designed to strengthen the information expression ability of key areas with the help of multi-scale feature extraction and attention mechanism. Finally， the lightweight deployment is achieved by model pruning and quantification technology. The results show that this algorithm performs well in the crack detection task，with an accuracy rate of 91. 4% and a recall rate of 87.3% ，which are 22.17% and 18.55% higher than the mainstream algorithms respectively. Moreover，the model reasoning time is only 8O milliseconds，and the parameter quantity is reduced to 12.3Mb . The research shows that the robustness and detail capture ability of the algorithm are obviously improved in complex background.

Key words： mold crack detection； computer vision； deep convolution generation countermeasure network （DCGAN）;convolutional neural network （CNN） ; multi-scale feature extraction

0 引言

模具是现代工业制造的核心工具，广泛用于汽车、航空航天、电子设备及家电等领域。（剩余10199字）