For classifying unknown 3-D objects into a set of predetermined object classes, a part-level object classification method based on the improved interpretation tree is presented. The part-level representation is implem...For classifying unknown 3-D objects into a set of predetermined object classes, a part-level object classification method based on the improved interpretation tree is presented. The part-level representation is implemented, which enables a more compact shape description of 3-D objects. The proposed classification method consists of two key processing stages： the improved constrained search on an interpretation tree and the following shape similarity measure computation. By the classification method, both whole match and partial match with shape similarity ranks are achieved; especially, focus match can be accomplished, where different key parts may be labeled and all the matched models containing corresponding key parts may be obtained. A series of experiments show the effectiveness of the presented 3-D object classification method.展开更多
A new approach for three dimensional (3-D) shape measurement was proposed based on colorcoded fringe and neural networks. By applying the phase-shift technique to fringe projection, point clouds were generated with hi...A new approach for three dimensional (3-D) shape measurement was proposed based on colorcoded fringe and neural networks. By applying the phase-shift technique to fringe projection, point clouds were generated with high spatial resolution and limited accuracy. The picture element correspondence problem was solved by using projected color-coded fringes with different orientations. Once the high accurate corresponding points were decided, high precision dense 3-D points cloud was calculated by the well trained net. High spatial resolution can be obtained by the phase-shift technique and high accuracy 3-D object point coordinates are achieved by the well trained net, which is not dependent on the camera model and will work for any type of camera. Some experiments verify the performance of this method.展开更多
基金The National Basic Research Program of China(973Program)(No2006CB303105)the Research Foundation of Bei-jing Jiaotong University (NoK06J0170)
文摘For classifying unknown 3-D objects into a set of predetermined object classes, a part-level object classification method based on the improved interpretation tree is presented. The part-level representation is implemented, which enables a more compact shape description of 3-D objects. The proposed classification method consists of two key processing stages： the improved constrained search on an interpretation tree and the following shape similarity measure computation. By the classification method, both whole match and partial match with shape similarity ranks are achieved; especially, focus match can be accomplished, where different key parts may be labeled and all the matched models containing corresponding key parts may be obtained. A series of experiments show the effectiveness of the presented 3-D object classification method.
基金Supported by the Eleventh Five-Year Pre-Research Project of China
文摘A new approach for three dimensional (3-D) shape measurement was proposed based on colorcoded fringe and neural networks. By applying the phase-shift technique to fringe projection, point clouds were generated with high spatial resolution and limited accuracy. The picture element correspondence problem was solved by using projected color-coded fringes with different orientations. Once the high accurate corresponding points were decided, high precision dense 3-D points cloud was calculated by the well trained net. High spatial resolution can be obtained by the phase-shift technique and high accuracy 3-D object point coordinates are achieved by the well trained net, which is not dependent on the camera model and will work for any type of camera. Some experiments verify the performance of this method.
