Classification Techniques Based On Reverse Engineering Features Of The "body"

There is a gap between reconstructed model of reverse engineering and modern CAD/CAM software, because the idea of reconstructed model based B-rep is different from modeling based feature. Reconstructed model based units is proposed in this dissertation. The key and core of this new idea is how to decompose the triangle mesh model and how to classify the features and to recover the designing idea. The unit recognition methods base on feature is unable to classify units which was blended, grooved, hit, wore, or operated by Boolean calculation, for they are limited to facet.After splitting the triangle mesh model of real part into units, classifying the feature units were studied in this dissertation. The major research contents and achievements of this dissertation are as follow:1. Reconstructed model based units in reverse engineering. The idea of modeling based feature is as follow: firstly, the part was split into designing features, and then modeled. On the contrary, the triangle mesh model was split into units firstly. Then, by classifying the units and recovering the design idea, the CAD model of real part was modeled. Both of these two ideas are consistent.2. Classifying feature units. A new method based shape distributions was developed. For the special data structure of triangle mesh model in reverse engineering, some random points were generated from the triangles firstly. Secondly, calculating the Euclidean distance between any two random points. Thirdly, constructing the shape distribution histograms (curve). Finally, comparing the histograms by EMD distance function to classify the different kinds of units. This method is easy to understand and insensitive to the shape change of feature units. A new method base on the continuity of shape distribution to get the feature parameters of cone by interpolation was proposed in this dissertation.3. Normalizing units. Minimum bounding box is applied in this process. The drive of normalizing units is that the same feature unit will has a different shape distribution for different modeling parameters. A method of normalizing units is proposed in this dissertation, which does well to the classifying units. This process is verified useful and valid during the experience.All of the algorithms and techniques presented in this dissertation have been realized and tested.