摘要
针对矿物浮选过程中泡沫图像处理的精矿品位建模存在有效泡沫图像样本缺乏、模型检测精度不足、泛化能力和鲁棒性较差等问题,提出了一种基于自注意力机制的变分自编码生成对抗网络(SA-VAEGAN)模型。其中,生成器使用由编码器和解码器组成的变分自编码器,编码层引入自注意力机制使卷积操作能更好地捕捉长距离依赖,获取全局信息,生成高质量的图像;判别器中嵌入分类器使其不仅有判别真假的功能,还能实现检测的目的。试验结果表明,该模型与其他检测模型相比有较强的泛化能力和鲁棒性,在精矿品位检测中准确率达到了96.67%。
Aiming at the problems of lacking effective froth image samples,insufficient model detection accuracy,as well as poor generalization ability and robustness in the modeling of concentrate grade in froth image processing during mineral flotation,a Variational Auto-Encoder-Generation Adversary Network based on Self-Attention mechanism(SA-VAEGAN)model was proposed.In this model,the generator employed the variational auto-encoder consisting of an encoder and a decoder,and the coding layer introduced a self-attention mechanism so that the convolution operation could better capture long-distance dependencies,obtain global information,and generate high-quality images.The classifier embedded in the discriminator not only had the function of discriminating true and false,but also achieved the purpose of detection.The test results show that the model has a strong generalization ability and robustness compared with other detection models,and the accuracy in concentrate grade detection reaches 96.67%.
作者
林俊
何港华
蔡耀仪
黄佳炜
LIN Jun;HE Ganghua;CAI Yaoyi;HUANG Jiawei(College of Engineering and Design,Hunan Normal University,Changsha,Hunan 410081,China)
出处
《矿业研究与开发》
CAS
北大核心
2023年第1期177-183,共7页
Mining Research and Development
基金
国家自然科学基金青年科学基金项目(61903138)
湖南省研究生科研创新项目(CX20200542)。
关键词
浮选
泡沫图像
品位检测
生成对抗网络
自编码器
Flotation
Froth image
Grade detection
Generative Adversarial Network
Auto-encoder