一种新型融合滤波器和CNN模型的仪表辨识方法

A New Instrument Identification Method Combining Filter and CNN Model

本项目以实际工程应用为背景,以数字仪表为研究对象,研究一种新的数字仪表辨识方法。本文介绍了CNN探测模式,实现了仪表数字盘的快速定位。针对已确定的仪表图像,提出了一种基于干扰捕捉的方法来标记可疑的干扰源,并对其进行有目的的处理。在此基础上,对过滤过程进行粒度划分,从而避开了对过滤限制的选取。同时,为了解决非均匀照明下的读数问题,本项目拟研究一种改进的单参量同态滤波方法,实现快速滤波与降低系统的参数调整数量相结合,大大降低计算速度。试验证明,这种算法可以实现多种场合下数字游标的精确辨识。

This project takes practical engineering applications as the background and digital instruments as the research object to study a new method for identifying digital instruments.This article introduces the CNN detection mode,which achieves rapid positioning of the instrument digital panel.A method based on interference capture is proposed to label suspicious interference sources for identified instrument images and perform purposeful processing on them.On this basis,the filtering process is divided into granularity to avoid the selection of filtering restrictions.At the same time,in order to solve the reading problem under non-uniform illumination,this project plans to study an improved single parameter homomorphic filtering method,which combines fast filtering with reducing the number of system parameter adjustments,greatly reducing calculation speed.Experiments have shown that this algorithm can achieve precise identification of digital cursors in various scenarios.