基于图像处理的二维位移测量系统

2-D displacement measurement system based on image processing

为改善现有图像位移测量,开发了一种二维位移测量系统。该系统采用3种非线性映射标定方法(BP神经网络、最小二乘(LS)三次曲面拟合、LS与遗传算法(GA)的组合方法(GALS)),基于Microsoft Visual Studio的集成平台。用该系统进行了位移测量实验,并分析了相关影响因素。结果表明:在这3种标定方法中,LS法的速度快、准确性高,综合性能最好;较低像素图像(不小于408×306)和较大的平滑、开运算模板(分别不大于标志点直径的1/2和1/3)有利于获得较好的测量效果。该系统可实现灵活、准确和实时位移测量。

A 2-D image-processing displacement-measuring system was developed to provide improved measurements.The system uses nonlinear mapping calibration algorithms,such as a back propagation（BP） artificial neural network（ANN）method,least square（LS） method,and an combined algorithm（GA-LS） from the genetic algorithm（GA） and LS,in a Microsoft Visual Studio integrated development environment.The system was used to analyze 他 the factors influencing displacement measurements.The results show that the cubic surface fitting based on the LS method is the best among the 3 calibration algorithms.Better measurements can be obtained with a smaller image（but larger than 408×306 pixels）,larger templates for smoothing（but less than 1/2 of a marking point diameter） and for opening operations（but less than 2/3 of the marking point diameter）.The measuring system is flexible and accurate and can be used in real-time.