No abstract

Semantic segmentation by using remote sensing images is an efficient method for agricultural crop classification. Recent solutions in crop segmentation are mainly deep-learning-based methods, including two mainstream architectures: Convolutional Neural Networks (CNNs) and Transformer. However, these two architectures are not sufficiently good for the crop segmentation task due to the following three reasons. First, the ultra-high-resolution images need to be cut into small patches before processing, which leads to the incomplete structure of different categories’ edges. Second, because of the deficiency of global information, categories inside the crop field may be wrongly classified. Third, to restore complete images, the patches need to be spliced together, causing the edge artifacts and small misclassified objects and holes. Therefore, we proposed a novel architecture named the Coupled CNN and Transformer Network (CCTNet), which combines the local details (e.g., edge and texture) by the CNN and global context by Transformer to cope with the aforementioned problems. In particular, two modules, namely the Light Adaptive Fusion Module (LAFM) and the Coupled Attention Fusion Module (CAFM), are also designed to efficiently fuse these advantages. Meanwhile, three effective methods named Overlapping Sliding Window (OSW), Testing Time Augmentation (TTA), and Post-Processing (PP) are proposed to remove small objects and holes embedded in the inference stage and restore complete images. The experimental results evaluated on the Barley Remote Sensing Dataset present that the CCTNet outperformed the single CNN or Transformer methods, achieving 72.97% mean Intersection over Union (mIoU) scores. As a consequence, it is believed that the proposed CCTNet can be a competitive method for crop segmentation by remote sensing images.

Radar detection is an advanced method for monitoring a blast furnace's inner burden surface shape, which is an important factor that largely affects the production efficiency of the iron-making process. In this paper, a radar detection-based model for the prediction of burden surface shape was developed for assisting operators in developing a charging strategy. The data used are composed of both the detection and controlling records of a real, working-state blast furnace obtained by mechanical swing radar and a furnace database system, respectively. By defining and analyzing the stacking density function, the physical meanings of the modeling principles were revealed. Combined with the classical force charging trajectory sub-model and detection-driven burden descent calculation, the proposed model adopts Gaussian radius basis functions to approximate the stacking mechanism of the burden charging process. The parameter identification results show that the model can approximate the burden surface radius profile well. Compared with the results obtained for coke layers, the parameters' ranges for the ore layers are narrower. Performance comparison shows that the proposed model has the advantages of higher prediction accuracy for both local details and global shape over the classical polygonal line model.

No abstract

The burden distribution process is an important and efficient measure to maintain the stable operation of the blast furnace. An accurate burden distribution model will reveal the impact on the internal furnace state and help to optimize the blast furnace production index. This article reviews the recent development of the modeling and control techniques in the burden distribution process. The current modeling methods of the blast furnace burden distribution can mainly be divided into the following types: the mechanismbased method, the physical scale model-based experiments and the data-driven method. However, most of the existing modeling methods are not applicable to general blast furnaces because it depends on the specific furnace structure and parameters. Furthermore, with the advancement in measurement technology, sensors now provide rich amount of online measurement of the blast furnace iron-making process. This makes the data analysis more challenging. It is imperative to establish new modeling methods for the burden distribution process. Therefore, this paper points out the new trends in modeling and control of the blast furnace burden distribution process. First, a dynamic clustering method based on dynamic time warping and adaptive resonance theory is introduced. Second, the inverse dynamic model-based burden distribution control is developed. Furthermore, a multi-model-based switch for modeling the fluctuating blast furnace process is formulated. Finally, the reinforcement learning method for the dynamic optimization of the production index is recommended.