卷积神经网络在矿石识别中的应用

Application of Convolution Neural Network in Ore Recognition

随着科技的进步,人工智能技术越来越多地应用在生活中的各个领域。在图像识别领域,计算机视觉技术和深度学习相关理论使得矿石分类成为可能。本文利用基于深度学习系统的pytorch框架下的卷积神经网络模型对七类矿石的手标本图像数据进行识别与分类,具有所需数据集小和训练速度快等优点。通过对矿石图像数据集进行划分,使用自定义卷积神经网络模型和经典神经网络模型VGG16对数据集进行训练,进而对预测结果进行对比分析。结果表明,自定义神经网络模型的准确率达到了76%,VGG16神经网络模型的准确率达到了87%。基本实现了矿石分类的目的。

With the advancement of science and technology,artificial intelligence technology is more and more applied in various fields of life.In the field of image recognition,computer vision technology and deep learning related theories make ore classification possible.In this paper,the convolutional neural network model under the pytorch framework based on the deep learning system is used to identify and classify the image data of hand specimens of seven types of ores,which has the advantages of small data set and fast training speed.By dividing the ore image dataset,the dataset is trained using a custom convolutional neural network model and a classical neural network model Visual Geometry Group Network 16,and the prediction results are compared and analyzed.The results showed the accuracy of the user-defined neural network model reached 76%,and the accuracy of the vgg16 neural network model reached 87%.The purpose of ore classification is basically realized.