The Study On Object Recognition Technology Based On Multispectral Polarization Imaging

The technology that can effectively recognize objects from complex backgrounds plays a significant role in object detection.One of the major problems facing China today is that the oil spill cannot be identified in the marine environment.The information obtained by single imaging technology is less and has more limitations.As an effective method of target detection,multispectral polarization imaging technology can not only distinguish the polarization difference between the target and the background,but it also can enhance the details of the image and improve the accuracy of object recognition in complex backgrounds by polarization information carried by itself and the intensity information obtained.In this paper,taking the oil spill as the experimental object,the new optical imaging theory and method of target recognition were studied,and the object recognition technology based on multispectral polarization imaging was proposed.The theoretical research and experimental verification were carried out by using the multispectral and multi-dimensional optical characteristics of polarization imaging.Firstly,object recognition was studied by obtaining the polarization characteristics of the object.Based on the theory of polarization degree,polarization phase and ellipticity,the polarization image of multi-dimensional information such as light intensity and polarization is obtained by the polarization imaging system.In order to make the gray image more convenient for human eyes to observe,the image was mapped to HSV space,and the gray image was converted to a color image.The gray value of the image was extracted to calculate polarization parameters.The results show that the boundary between object and background can be recognized by different colors.Different oil products can be distinguished clearly according to different parameters.Secondly,in this paper,the quaternion theory of polarized light was introduced to explore the influence of the reflection characteristics of polarized light on object recognition.The quaternion was combined with the Fresnel formula,and the quaternion formula of Stokes and Poincare sphere were derived,which makes the amount of calculation greatly reduced compared with the calculation of polarization optical parameters by the Mueller matrix.According to the Fresnel formula of quaternion form,the relationship between quaternion component and polarization state of incident light was simulated by MATLAB software.The s component,p component,and quaternion of reflected light of different oil products were calculated through experiments.The calculation results show the numerical difference among each oil product is obvious.By combining the quaternion sphere with the Poincare sphere,it is observed that the polarized state of reflected light of the object is different in the spherical surface,so it can be more intuitive to identify different objects.Finally,the object identification under multi-wavelength was studied by estimating the refractive index of the object by using the multispectral polarization image.Cauchy dispersion equation and refractive index calculation principle were reasonably used in the algorithm.Combined with the experiment,the refractive indexes of the metals used in the experiment were complex,and the refractive index of the oil film was between 1.3-1.5 at a specific wavelength.Through the experimental dispersion coefficient image representing material properties were fused and mapped to the color space,object recognition can be realized according to different image colors.Besides,the position of the quaternion of the reflected light on the spherical surface under multi-wavelength was further explored.The experimental results show that the position of each object changes with the wavelength under the effect of different detection light wavelengths,but does not deviate greatly.