Font Size: a A A

Research On 3D Measurement For Jaw Based On Multiple Views Geometry

Posted on:2017-10-20Degree:MasterType:Thesis
Country:ChinaCandidate:C Y GuoFull Text:PDF
GTID:2348330509962972Subject:Mechanical Manufacturing and Automation
Abstract/Summary:PDF Full Text Request
The jaw 3D measurement is an important step in the digital dental diagnostic. Compared with the traditional dental mask technology, the digital jaw 3D measurement is better. In order to solve the limited space of the high precision three dimension measurement for jaw and the special requirement of convenience, this paper proposes the method with multiple views geometry and realizes the rapid measurement with a single camera snapping freely. The main research contents are as follows:(1) To realize the dynamic dental measurement, a portable dental measurement model based on the multiple views geometry oriented to the special request of the jaw model is designed and verified. The system parameters are accurately calibrated using Zhang’s method. And the three dimension measurement for jaw based on the single camera snapping freely is realized. To solve false match problem caused by the not obvious feature on jaw surface under the little view, SIFT matching algorithm is improved. Using the Hellinger kernel to match the key points’ descriptors, the image matching algorithm is optimized. The feature points’ 3D information is recovered and the sparse reconstruction is implemented.(2) To improve the stability of the essential matrix estimation, a new robust, accurate method based on the optimized RANSAC is proposed. The experiment shows that the method can effectively improve the reconstruction accuracy of the two views. To lower the accumulate error brought out by the multiple images, the camera pose is calculated precisely using the scale factor of the point sets. And then the single piece of the dense point cloud is reconstructed.(3) The double neighborhood alignment algorithm based on the fast point feature histograms is proposed, and the multiple point sets registration in space is achieved. Several sets of experimental data demonstrate that the method can effectively handle the point cloud with their character not being obvious, and satisfy the requirement of the accuracy and stability of the dental measurementBased on above the research results, the portable jaw measurement system based on the multiple views geometry is developed. The function modules of the system calibration, feature extracting, sparse reconstruction, dense reconstruction and the point cloud registration etc. are initially achieved and the experiment results are good.
Keywords/Search Tags:3D measurement, multiple views geometry, image matching, essential matrix, point cloud registration
PDF Full Text Request
Related items