基于CR下界无偏能量估计的高炉料面点云锐化成像

高炉雷达获取的料面信息是钢铁冶炼中布料控制的重要参数.但高炉内部环境复杂, 料面具有非均匀流态化特性, 传统信号处理方法难以准确稳定提取料面有效信息, 会导致高炉布料误操作.本文借鉴遥感SAR雷达成像原理, 设计了工业SAR扫描式雷达, 多倍增加料面采样点密度, 提出一种新的料面点云锐化成像算法. 分析了高炉雷达料面回波信号干扰信号特征, 从图像处理角度, 设计多级滤波器对2D频谱图进行去噪处理分离出一条带状的料面回波信号区域. 对料面距离估计问题, 基于克拉美罗下界(Cramer-Rao lower bound, CRLB)提出一种先加权采样锐化料带峰脊再利用能量重心法估测料面距离频率的方法, 生成3D料面点云模型, 并利用CRLB评价本文算法性能.在恶劣条件下, 实测高炉雷达料面回波信号的点云成像验证显示, 本文方法优于传统寻峰法, 能有效处理低信噪比信号, 准确提取料面有效信息.同时料面距离频率估计精度更高, 且相较于其他方法频率估计误差更接近CRLB下界.

Sharpness Image of Burden Point Cloud Based on CR Lower Bound Unbiased Energy Estimation

Image information obtained by the radar is an important parameter in the control process of the distribution in iron and steel smelting. The interior of the blast furnace presents high complication. The traditional signal processing method is difficult to extract the effective information accurately and stably, which seriously misleads blast furnace charging operation. Based on the imaging principle of remote sensing SAR radar, this paper designs the industrial SAR scanning radar, which multiplies the density of the sampling points of the surface. We propose a new image forming algorithm of the point cloud surface. In this paper, the characteristics of the echo signal of the surface of the blast furnace radar are analyzed. We design multistage filters for 2D spectrum as a denoising operation in order to isolate a strip of surface echo signal area. To estimate the distance of the blast surface, based on the Cramer-Rao lower bound (CRLB), a new method is proposed to estimate the distance frequency of the blast surface by using energy centrobaric correction method after the weighted sampling of the peak ridge of the sharped surface belt, and the 3D mesh point cloud model is generated, and CRLB is used to evaluate the performance of the algorithm in this paper. In harsh conditions, the point cloud imaging verification of measured echo signal from blast radar shows that proposed method is superior to the traditional peak searching method, and it can effectively process the signal with low signal-to-noise ratio and accurately extract the effective information of the material surface. At the same time, the accuracy of frequency estimation is higher, and the error of frequency estimation is closer to CRLB lower bound than that of other methods.