摘要
在分析煤矸石分拣环境特点及煤矸石视觉特点的基础上,提出改进的卷积神经网络煤矸石图像识别算法,并从损失函数、模型参数以及准确率3个方面进行分析研究。结果表明:改进后的卷积神经网络图像识别算法能有效地避免分选环境中的噪声影响;与传统的分选方法相比,具有更快的识别速度和更高的准确率,能更好地满足实际工程需要。
Based on the analysis of the characteristics of gangue sorting environment and gangue vision,an improved convolution neural network algorithm for gangue image recognition was proposed,and the loss function,model param eters and accuracy were analyzed.The results show that the improved convolutional neural network image recognition algorithm can effectively avoid the impact of noise in the sorting environment;compared with the traditional sorting method,it has faster recognition speed and higher accuracy,and can better meet the actual engineering needs.
作者
何克焓
HE Kehan(School of Electromechanical and Information Engineering,China University of Mining and Technology(Beijing),Beijing 100083)
出处
《河南科技》
2020年第4期66-68,共3页
Henan Science and Technology
关键词
煤矸石
卷积神经网络
图像识别
损失函数
coal and gangue
convolutional neural network
image recognition
loss function
