基于遗传卷积神经网络的微小零件缺陷分类

Classification of Small Part Defects Based on Hybrid Genetic Convolutional Neural Network

针对微小零件分类的复杂性,以及各种干扰和个体差异现有分类方法精确度较低的问题,提出利用一种基于遗传卷积神经网络(GA-CNN)。该网络具有分类准确率高、模糊性好的特点。首先利用高清相机采集图像,然后对图像进行预处理,最后利用GA-CNN网络对图像进行分类。通过实验表明该网络分类准确率相比去单一的卷积神经网络准确率高,比较适合在现代的智能制造零件检测中。

Aiming at the complexity of micro-parts classification and the low accuracy of existing classification methods based on interference and individual difference,a new method based on genetic convolutional neural network(GA-CNN)is proposed.The network has the characteristics of high classification accuracy and good modularity.First,the high-definition camera is used to collect the images,then the images are preprocessed,and finally the images are classified by GA-CNN network.Experiments show that the network classification accuracy rate is higher than the single convolutional neural network,more suitable for modern intelligent manufacturing parts detection.