摘要
针对传统选煤方法在煤矸识别上效率低、错选率高等问题,提出了视觉图像加卷积神经网络的煤和矸石识别新方法。在团队研发的煤矸分拣机器人平台上采集了煤矸图像数据并进行了扩充处理。以卷积神经网络VGG_16为基础设计改进了模型,通过设置不同的模型参数验证了其在煤和矸石识别上的性能。结果表明,新模型能在占用很少的硬件资源下达到较高的煤矸识别率,当网络学习率设置为0.0001和正则化系数设置为0.001时模型的性能达到最优,训练集和测试集的识别准确率分别达到了99.73%和97.58%。
Aiming at the problems of low efficiency and high error separation rate of traditional coal separation methods in coal and gangue identification,a new method of coal and gangue identification based on visual image and convolutional neural network was proposed.The image data of coal and gangue were collected on the coal and gangue sorting robot platform developed by the team and expanded.Based on convolutional neural network VGG_16,the model is improved and its performance in coal and gangue identification is verified by setting different model parameters.The results show that the new model can achieve a high identification rate of coal and gangue with less hardware resources.When the network learning rate is set to 0.0001 and the regularization coefficient is set to 0.001,the model achieves the best performance,and the recognition accuracy of training set and test set can reach99.73%and 97.58%,respectively.
作者
李亚坤
马宏伟
王鹏
LI Yakun;MA Hongwei;WANG Peng(School of Mechanical Engineering,Xi′an University of Science and Technology,Xi′an 710054,China;Shaanxi Key Laboratory of Intelligent Monitoring for Mine Mechanical and Electrical Equipment,Xi′an 710054,China)
出处
《煤炭技术》
CAS
北大核心
2022年第9期156-159,共4页
Coal Technology
基金
国家自然科学基金面上项目(51975468)。
关键词
煤矸识别
卷积神经网络
数据扩充
改进VGG_16网络
coal and gangue recognition
convolutional neural network
data expansion
improved VGG_16 network