Research on structure semantic automatic annotation for 3D model
This project is mainly focusing on 3D model segmentation and structure semantic automatic annotation algorithm. It is aiming to achieve consistent segmentation result by using semantic knowledge constrain. In addition, the semantic features of each part can be identified automatically. This thesis takes the human body as example, and generates a framework of human semantic analysis based on the domain knowledge.
The proposed framework includes two sides, one is the formalization semantic knowledge of human body; the other is human automatic segmentation and semantic analysis algorithms by using defined domain knowledge. The ontology and RDF language are used to represent semantic knowledge of 3D human.
The whole semantic analysis algorithm for human body includes the following steps.
The results show that the algorithm can avoid the inconsistency problem very well. Furthermore, it can identify semantic feature points accurately for human body.
The proposed framework includes two sides, one is the formalization semantic knowledge of human body; the other is human automatic segmentation and semantic analysis algorithms by using defined domain knowledge. The ontology and RDF language are used to represent semantic knowledge of 3D human.
The whole semantic analysis algorithm for human body includes the following steps.
- It calculates the Poisson signal value of 3D model and extract outside key points based on geodesic distance;
- It uses the extracted key point as seed point to perform a parallel region growing process. The process is to get the coarse segmentation results;
- According to coarse segmentation results, the algorithm can identify part semantic and human semantic feature points by using the human domain knowledge and Poisson signal values.
- The graph cuts theory to optimize the segmentation boundaries with the help of human semantic feature points. In experiments, different posture of the human body models is tested to verify the effect of the algorithm.
The results show that the algorithm can avoid the inconsistency problem very well. Furthermore, it can identify semantic feature points accurately for human body.
a_hierarchical_model_for_3d_human_feature_point_extraction.pdf | |
File Size: | 3386 kb |
File Type: |
C# Tank game
It's a simple game with C#, the environment is VS2010 and it's a very old game. You can choose single player or two players. There is also a map editor, choose brick, grass, water and so on. Save the map then it can be used in the game. There is basic shooting and hit testing in it. It helps me to get familiar with C# quickly.
Android Mp3 Player
This is a Mp3 Player which is used for me to get to know the android programming. This mp3 player can check songs on the web site, and choose to download songs you like in the local SD card and can play script while playing the song.There is a song list on the web site to store the list of songs which is an XML file. The local player connect to the XML file and shows the songs on the website. Then user can choose to download them. Since I'm not an art, the image style looks ugly. But all the basic function works well.