Model-oriented 3D Object Detection And Reconstruction Based On RGB-D Video

In recent years,3D reconstruction has received more and more attention from scholars in the field of computer vision.As the requirements for 3D reconstruction become higher and higher,the difficulty becomes more and more obvious.The difficulty of 3D reconstruction lies in the following reasons: the range of the object's movement,the accuracy of the 3D point groups affected by noise,the mutual occlusion of the point groups,the speed of the 3D reconstruction,and other factors.In the reconstruction methods proposed in recent years,the image data is captured in a simple scene or a target is placed on a rotating platform.So in order to reduce the limitation of the scene,in this paper,we explore how to use the same type of 3D model as the target to assist the user to detect the pose and position of the target and achieve the purpose of reconstruction.In this paper,a single RGB-D camera is used to obtain 3D images,and a model-oriented 3D object detection and reconstruction method based on RGB-D video is proposed.The use of RGB-D image data is rich in information and effectively increases the accuracy of target detection and attitude estimation.Utilizing the RGB-D data of different perspectives provided by the three-dimensional target samples established in advance,the target of the input image can be accurately detected,and the target pose estimation can be effectively performed when the internal and external parameters of the camera cannot be known.With improved tracking and alignment technology,this paper can fulfill the needs of 3D image display and scene reconstruction for augmented reality applications.Compared with traditional pose estimation using a small number of features,this method can reconstruct high-quality 3D objects more efficiently.In order to verify the reliability of the reconstruction system,the method used in this paper is compared with other attitude estimation methods,and the results show that this method has obvious advantages,proving that the reconstruction system can effectively detect the target and attitude estimation.