基于光斑识别的角度传感器及其非线性补偿技术  

Angular sensor and its nonlinear compensation technology based on spot recognition

作　　者：肖光华 侯庆民 张靖 李鑫 贾子光 XIAO Guanghua;HOU Qingmin;ZHANG Jing;LI Xin;JIA Ziguang(College of Architectural Engineering,Chengdu Vocational&Technical College of Industry,Chengdu 610213,China;School of Physics and Information Engineering,Zhaotong University,Zhaotong 657000,China;College of Marine Science and Technology,Dalian University of Technology,Dalian 116024,China)

机构地区：[1]成都工业职业技术学院建筑工程学院,四川成都610213 [2]昭通学院物理与信息工程学院,云南昭通657000 [3]大连理工大学海洋科学与技术学院,辽宁大连116024

出　　处：《传感器与微系统》2025年第1期97-101,共5页Transducer and Microsystem Technologies

基　　金：国家重点研发计划资助项目(2019YFB1504303);成都工业职业技术学院大师工作室(XJDS-2022-1)。

摘　　要：为了解决现有霍尔角度传感器存在成本高、易受电磁干扰的问题,设计了一种基于光斑识别的角度传感器,将摄像头与激光发射器放置在一个密封的铝盒内,由旋转轴带动激光发射器转动,并向摄像头发射激光束,在摄像头端采集到的图像上形成不同像素的光斑,将每一个光斑中心点坐标转换为旋转轴转动的角度。采用基于最小二乘的椭圆拟合算法,对光斑中心点像素坐标进行计算。再利用高斯曲线拟合法,对输入输出曲线进行拟合,以分析传感器的非线性误差。并利用反向传播(BP)神经网络算法对传感器进行非线性补偿。实验结果表明:经BP神经网络算法补偿后,传感器的平均非线性误差为1.63、均方根误差为1.24,比未补偿前分别降低了0.61和2.56。且测量的均方误差为0.8°。To address the issues of high cost and susceptibility to electromagnetic interference of existing Hall angle sensors,a novel anuglar sensor based on spot recognition is designed.The camera and laser emitter are housed within a sealed aluminum box,the laser emitter rotated by a rotational axis to project a laser beam towards the camera.This results in the formation of light spots with different pixel coordinates on the captured image by the camera,which are then converted into the angular position of the rotational axis.The light spot center point pixel coordinates are transformed to angle of rotation,and calculated using an ellipse fitting algorithm based on the least squares method,and the input-output curve is fitted using the Gaussian curve fitting method to analyze the nonlinear error of sensor.Furthermore.Back propagation(BP)neural network algorithm is employed for nonlinear compensation of the sensor.Experimental results demonstrate that after compensated by BP neural network algorithm the sensor average nonlinear error and the root mean square error is reduced to 1.63 and 1.24,respectively,a decrease of 0.61 and 2.56,respectively,compared to the uncompensated state.The measured mean square error is 0.8°.

关 键 词：角度传感器 光斑识别 最小二乘的椭圆拟合 高斯曲线拟合 反向传播神经网络 

分 类 号：TP212[自动化与计算机技术—检测技术与自动化装置] O439[自动化与计算机技术—控制科学与工程]