基于深度学习的模具裂缝智能检测与图像分割算法研究

打开文本图片集

中图分类号：TG241 文献标志码：A

Abstract： Mold crack detection is an important link to ensure product quality and production safety in industrial manufacturing. The purpose of this paper is to propose an intelligent detection and image segmentation algorithm for mold cracks based on deep learning（DL） to solve the problem of crack identification in complex background. Firstly，a high-quality data set of mold cracks was constructed，and then the convolution neural network （CNN） model was improved by multi-scale feature fusion strategy to enhance its ability to detect cracks of different sizes. The results show that the accuracy of this algorithm is 92.65% and the recall rateis 89.73% . Compared with the traditional threshold segmentation method，it is improved by 21.51% and 19.07% respectively. The IoU reaches 86.47% ，and the inference time is only 32.89ms ，which can support real-time detection at 3O.41 frame·s-1 ．The research shows that the algorithm is superior to the existing methods in accuracy，robustness and calculation efficiency，and has certain industrial application value.

Key words： mold crack detection； deep learning； image segmentation； multi-scale featurefusion；real-time

0 引言

在现代工业生产中，模具是制造业的基础工具，广泛用于汽车、航空航天、电子设备及各类消费品制造。（剩余8397字）