Polarization Binocular Three-dimensional Imaging For The Suppression Of Metal Surface Flares

Metal target becomes the difficulty of 3D imaging because of its special surface optical characteristics.The metal target with smooth and highly reflective surfaces will produce flare,which often causes the lack of data on the target surface.In order to improve the 3D imaging accuracy and detail of such objects,a binocular stereo vision 3D imaging technique combining polarization is studied in this thesis.This thesis analyzes the shortcomings of traditional binocular vision in metal target imaging.Then,the structured light of random speckle matrix is introduced to generate artificial feature points.Based on this,the imaging system and the feature point matching algorithm for calculating parallax are improved.It is concluded that the main factor affecting metal imaging is specular reflection light by comparing the imaging accuracy of diffuse reflection target and metal target.By analyzing the causes of metal surface flares and partially polarized light,it is proposed to suppress flare by using polarizers.In the experiment,the single polarization point cloud fusion method and the orthogonal polarization image fusion method are respectively used for 3D imaging of metal targets.In addition,the 3D imaging method of metal surface based on polarization state information is studied,and the calculation algorithm of polarization state,zenith angle and other parameters is realized by MATLAB program,and the simple 3D surface imaging based on these parameters is also simulated.The experimental results show that the metallic surface flare caused by specular reflection makes the dispersion of 3D point cloud increase sharply.When the working distance is 1m,the standard deviation of plane point cloud increases from 1.04 mm to 2.95 mm,and the recognition rate of speckle is only 37.53%.However,the standard deviation of polarization binocular imaging is 2.31 mm,the average deviation is 1.78 mm,the absolute error between the fitting plane and the actual plane is less than 1mm,and the point recognition rate is 77.48%.Therefore,the polarization image fusion method has a good suppression effect when the metal surface is highly flared,which can improve the 3D imaging accuracy and details of the metal target.But this method is not suitable for low flare imaging.When the target is low flared and in close range,it is more suitable to use the 3D imaging based on polarization information in theory,but it has higher requirements for light source.