Invariant Moments Of Illumination, Blur Invariance And3D Spacecraft Recognition Algorithm

Invariant moments is widely used in the aerospace, medical, marine, and even normal life,especially in spacecraft in the last few years.Therefore, this paper is based on the cognitive processesof the spacecraft, based on the invariant moments of illumination, blur invariance and3D spacecraftrecognition algorithm.First, duing to affine deformation and blur deformation to the space target image, space targetrecognition combine the affine invariant moments and the blur constant moment, to solve the problemof how different angle and different distance space target recognition.Second, in order to solve the invariant moments changing under the conditions of illuminationchanges, this paper proposes light-affine fusion invariant moments, based on gray light model andLambertian color model. Undering different lighting, angles and affine changes satellite, usingminimum classification criteria.The average recognition accuracy is93.71%, comparison the classicgray image illumination invariant moments improved by20%, and is suitable for a color image andgray image.Then, in order to solve the invariant moments changing under the changes of the illuminationand blur, a light-fuzzy fusion affine invariant moments are proposed and apply to the identification ofthe spacecraft, combined with light-affine fusion invariant moments and blur invariant model. Theaverage recognition accuracy is94.60%, comparison the Jan affine invariant moments improved by17.34%.It suitable for the different angles and different light-blur target recognition problem,enhancing target recognition robustness.Third, in order to solve the high computing speed of the target image moments, the paper use IIRdigital filter algorithm to calculate the target image moments to reduce the invariant momentscomputing time,with analysis of moment invariants fast algorithm and the feature of spacecraftlight-blur transformation.Finally, in order to solve the three-dimensional target transformed into two-dimensional imagerecognition expend more times. The paper gives a moment invariants and PCA fusion method basedon the principal of the PCA, improves real-time17.8%. And light-blur fusion combinations ofinvariant moments with PCA fusion apply to the identification of the spacecraft. Probabilistic NeuralNetwork (PNN) recognition accuracy is more than98%with three different satellite model for experiment, comparison classic combination invariant moments accuracy improved18.19%. Foradaping to different light, blur, affine transformation and real-time recognise.