危化品码垛仓储的定位数据中值滤波去噪算法

De-Noising Algorithm Based on Median Filtering for Positioning Data of Hazardous Chemicals Palletizing Warehousing

现有危化品仓储的堆垛安全监测主要是针对仓储堆垛的五距(堆距、墙距、顶距、柱距、通道距)的监测。实际应用时,根据"五距原则"将堆垛划分成多个检测面,采用堆垛激光扫描定位装置,实现对每个堆垛检测面的监控。结合各个面的监测数据,可以实现对仓库货位堆垛的整体安全监测。危化品堆垛检测的激光扫描定位装置所采集的数据中含有逸出、失落信号等奇异点数据。这些奇异点数据对码垛监控的精准度有一定影响。针对点云数据中的噪声数据特性进行了分析,采用中值滤波减少"逸出值"对拟合数据的干扰,借助多段最小二乘法实现对"失落点"的数据信息拟合,最终实现对点云数据的联合去噪。试验结果表明,基于中值滤波与多段最小二乘法能拟合出设备扫描截面的形状信息,有效降低了奇异点对定位监控设备的影响。

The existing safety regulatory for hazardous chemicals warehousing is mainly based on the five-distance（ stacking distance,distance from the wall,distance from the top,distance from the column and distance of aisles） regulation. In practical application,in accordance with the five-distance regulation,the stacking area can be divided into multiple monitoring surfaces;the monitoring of each detection surface of stack can be realized by laser scanning positioning device,and the overall safety inspection can be achieved for stacking in warehouse. The data collected by such device may contain singular points,e. g.,escaped and lost signals,etc.,and these data may affect the monitoring accuracy. The characteristics of these noise data in point cloud data collected are analyzed,and the interference of ＂escaped value＂on fitting data is reduced by using median filtering.With help of multi-stage least square method,the data fitting of ＂lost point＂is implemented; then combined de-nosing of the point cloud data is implemented at last. Experimental results show that the shape information of scanned cross section can be fitted with the method based on median filtering and multi-stage least square,and the influence of singular point on the positioning and monitoring equipment is reduced effectively.