Research On Improving The Accuracy Of Structured Light 3D Measurement Based On Image Processing

In recent years,the rapid development of modern manufacturing and other fields requires smaller and smaller measurement errors.As a typical high-precision and highefficiency measurement method,the structured light measurement method has received widespread attention in recent years.The accuracy of the structured light measurement system is based on the image quality.However,many factors in the measurement process degrade the quality of the stripes image,resulting in image distortion,blur,or reduced signal-to-noise ratio.These affect the computer's calculation of the threedimensional information of the object,and the measurement accuracy is reduced.Aiming at the degradation of structured light stripe image,this thesis carried out related research to improve the measurement accuracy.The specific work is as follows:Firstly,the basic principles of several three-dimensional measurement methods of structural light are introduced.The errors in images are analyzed systematically from the reasons of the degradation of stripes,including the projection and capture,the interference factors of the measured object,the environment,image data transmission and other aspects.Secondly,aiming at the image quality degradation of structured light stripes,this thesis proposes a set of stripe image quality evaluation schemes.This method uses the skewness coefficient defined based on the third-order center distance and the no reference image quality evaluation method,which is based on Human Vision System(HVS),to evaluate the quality of the stripe image.According to the relationship between the skewness coefficient and the positioning error of the stripe center,the skewness coefficient is used as the standard to evaluate the asymmetry degree of the stripe gray distribution.the regions of interest in the images are selected by simulating human visual characteristics.Calculating the noise variance by using the image block processing method,and the noise level in the image is estimated by using the filter quadratic blur and the variance calculation.The noise variance of the selected regions is used to replace the noise variance level of the whole image,and the evaluation value of the noise level of the stripe image is proposed.By setting the thresholds of these two parameters,the data with large errors can be identified and eliminated.The results of simulation and experiment verify the feasibility of the method.In addition,aiming at the noise interference in structured light stripe image,the common image filtering methods are analyzed,and an image filtering method based on the principle of liquid surface tension is proposed.In this method,the pixels in the target image are analogized to water molecules on the liquid surface.The compensation method of pixel gray value is given by simulating the intermolecular force and theirs force state,and the pixel gray value compensation model is established.The filtering method based on liquid surface tension principle is used to process the images.By optimizing the gray distribution of stripes,the noise is removed and the image quality is improved.Finally,the feasibility and effectiveness of the above method are verified by simulation and experiment.