【摘要】 为减少卷烟厂烟丝生产过程中人工检测烟柜状态带来的安全隐患，提出一种改进的卷积神经网络对烟柜状态进行自动识别。使用工业相机对不同烟柜的多种状态进行数据采集并做数据增强，构建烟柜状态数据集；设计具有17层卷积层、5层池化层、3层全连接层的改进的卷积神经网络模型，对烟柜状态进行识别。通过实验与图像分类经典模型作比较，实验结果表明，提出方法对不同烟柜的识别准确率已高达98.01%,优于现有方法，对烟厂的安全高效生产具有实际意义，对其它计算机视觉的识别问题也具有一定借鉴作用。更多还原

【Abstract】 To reduce the hidden danger caused by manual detection of the tobacco cabinet status in the process of cigarette production in cigarette factories, an improved convolution neural network was proposed to realize the automatic recognition of tobacco cabinet state in a cigarette factory. The industrial camera was used to collect and enhance the data of various states of different tobacco cabinets, and the tobacco cabinet state data set was constructed. An improved VGG convolutional neural network model with 17 convolution layers, five pooling layers, and three full connection layers was designed to identify the status of the tobacco cabinet. It was compared with the classic model of image classification through experiments. Experimental results show that the recognition accuracy of different tobacco cabinets reaches 98.01%, which is better than that of existing methods. The proposed method has practical significance for the safe and efficient production of the cigarette factory. It will provide a reference for other computer vision identification problems.更多还原