为了智能、高效分离选煤厂原煤中的杂物,促进选煤工艺智能化发展,提出了一种结合卷积神经网络(CNN)和Transformer的语义分割模型,专门针对振动筛上复杂环境中的杂物进行像素级分割。首先,将Transformer结构应用到选煤厂振动筛的杂物分割任务上,实验结果表明其能更准确地识别杂物边缘。其次,详尽对比分析了不同的上采样方法和Transformer模块,增强了模型的可解释性和适用性。最后,通过与其他先进模型(如ResNet和Swin-Transformer)的综合比较,证实了提出模型在处理振动筛上杂物分割问题上的优越性能。

In order to intelligently and efficiently separate impurities from raw coal in coal preparation plants, and promote the intelligent development of coal preparation processes, we propose a semantic segmentation model that combines Convolutional Neural Networks (CNN) and Transformers, specifically designed for pixel-level segmentation of tramp materials in complex environments on vibrating screens. Firstly, we pioneer the application of the Transformer structure for tramp material segmentation tasks on vibrating screens in coal preparation plants. Experimental results demonstrate its superior ability in accurately delineating the edges of tramp materials. Secondly, we provide a comprehensive comparison and analysis of various upsampling methods and Transformer modules, thereby enhancing the interpretability and applicability of the model. Finally, through an exhaustive comparison with other state-of-the-art models such as ResNet and Swin-Transformer, we validate the efficacy of the proposed model in handling the segmentation of tramp materials on vibrating screens.