Three-dimensional Target Plane Rotation To Identify

The research of three-dimensional object recognition plays an important role in industrial production and has become the significant research problem in engineering area. In this paper, we focus on the problem of three-dimensional object recognition under out-of-plane rotation and complex background. Theoretical research and experi ment have been carried out as the follows:In order to recognize the 3D object under out-of-plane rotation and nonlinear illumination, 180 with different rotation angle were taken as the set of training images. By using principal component analysis method, twenty feature vectors were selected and the dimension number of feature space is greatly reduced. The image decomposition and reconstruction were processed based on these feature vectors, and the complexity of calculation was decreased while keeping higher calculation accuracy. The object recognition rule based on the cosine value between original image vector and reconstructed image vector is proposed. Computer simulation results show, the method can not only effectively recognize the object with different rotation angle, but also eliminate the influence of nonlinear illumination.In order to recognize the object and determine its spatial position, 180 trained images are divided into four image sets, and the feature vectors of each image set are calculated. It is shown that each image set is represented with three feature vectors, and thus the total four feature spaces are constructed with only twelve feature vectors. As a result, the object decomposition and object reconstruction are easily achieved. Based on the relation between object vector and its reconstructed vector, we can not only determine whether the object is true or false, but also locate its position in space. Computational simulation results show that method of constructing multiple feature spaces and the object recognition rule are effective, the goal of recognizing the object and determining its spatial position under out-of-plane rotation is achieved.In order to recognize the 3D object under complex background, we use deformable templates to decrease the effects of complex background. The goal of recognizing the object under complex background is achieved.