基于Alexnet的金相识别研究

Research on Metallographic Identification Based on Alexnet

对《机械工程材料综合实验》课程后期的金相组织图像处理,以深度学习代替人眼观察进行自动识别分类。通过建立金相组织图像库,分析各类材料金相组织的不同形态特征,结合Alexnet神经网络模型,研究面向不同组织金相识别的Alexnet迁移学习方法,实现金相图像的自动识别。在4类不同的金相组织图像样本库上进行图像识别实验,准确率达到99.63%,能够快速地实现不同材料的区分。与通过人眼观察的方法对比,可大大减少人工操作的工作量,且方法迅速准确。

Automatic recognition and classification based on deep learning instead of human eye observation in the metallographic structure image processing of"comprehensive experiment of mechanical engineering materials"course.Through the establishment of metallographic structure image database,the different morphological characteristics of various materials'metallographic structures are analyzed.Combined with the Alexnet neural network model,the Alexnet migration learning method for metallographic identification of different structures is studied,and the automatic recognition of metallographic images is realized.The experimental results show that the accuracy of image recognition is 99.63%,which can quickly distinguish different materials.Compared with the method of human eye observation,it can greatly reduce the workload of manual operation,and the method is rapid and accurate.