Research And Implementation Of Pose Estimation For High Inclination Plane Targets

With the continuous development of the Internet and mobile terminals,AR(Augmented Reality)and VR(Virtual Reality)applications are gradually emerging.Pose estimation of plane target is one of the key technologies,and there are many research results.However,in practical applications,since the angle of the photographer’s screen is too inclined,the posture of the target object changes too much.Subject to the hardware conditions of people using equipment,these special situations are often not handled well.In the research of pose estimation,the processing methods for the special scene with a large target tilt include analyzing the texture features of the image,using a new feature description method,and enhancing the affine invariance of the image through affine simulation.However,there are many problems such as redundant computation and decreased accuracy in the process of pose estimation.In this paper,aiming at the special scene where the target is inclined to a large degree in pose estimation,an image matching pose estimation algorithm based on optimal radiation simulation is proposed,and the pose estimation results are optimized to improve the calculation efficiency and calculation accuracy in the high-inclined scene.In the feature matching of high-angle images,the redundant affine simulation process is avoided by selecting the optimal affine simulation in advance,and the accuracy of pose estimation is ensured by improving the super-descriptor calculation method of feature points.The point-to-point PnP(perspective-n-point)pose estimation algorithm optimizes the process of sampling and screening the initial solution,so that the results of attitude estimation can converge faster and meet the accuracy requirements faster.Finally,this paper designs a general process for plane marker pose estimation under high inclination,and implements a modular system that can be decomposed by multi-step algorithms.Each module can output its own intermediate steps,which is convenient for algorithm comparison and evaluation of experimental results.The experimental comparison shows that,compared with the traditional PnP attitude estimation algorithm based on full affine simulation,the method proposed in this paper can effectively improve the calculation speed of the plane target attitude estimation under high inclination angle and ensure its accuracy,which enhances the practical application of the algorithm and the application value in engineering.