In this paper, we present a novel approach for 3D dental model segmentation via deep Convolutional Neural Networks (CNNs). Traditional geometry-based methods tend to receive undesirable results due to the complex appearance of human teeth (e.g., missing/rotten teeth, feature-less regions, crowding teeth, extra medical attachments, etc.). Furthermore, labeling of individual tooth is hardly enabled in traditional tooth segmentation methods. To address these issues, we propose to learn a generic and robust segmentation model by exploiting deep Neural Networks, namely NNs. The segmentation task is achieved by labeling each mesh face. We extract a set of geometry features as face feature representations. In the training step, the network is fed with those features, and produces a probability vector, of which each element indicates the probability a face belonging to the corresponding model part. To this end, we extensively experiment with various network structures, and eventually arrive at a 2-level hierarchical CNNs structure for tooth segmentation: one for teeth-gingiva labeling and the other for inter-teeth labeling. Further, we propose a novel boundary-aware tooth simplification method to significantly improve efficiency in the stage of feature extraction. After CNNs prediction, we do graph-based label optimization and further refine the boundary with an improved version of fuzzy clustering. The accuracy of our mesh labeling method exceeds that of the state-of-art geometry-based methods, reaching 99.06 percent measured by area which is directly applicable in orthodontic CAD systems. It is also robust to any possible foreign matters on model surface, e.g., air bubbles, dental accessories, and many more.

3D Tooth Segmentation and Labeling Using Deep Convolutional Neural Networks

To solve the problem of low efficiency, the complexity of the interactive operation, and the high degree of manual intervention in existing methods, we propose a novel approach based on the sparse voxel octree and 3D convolution neural networks (CNNs) for segmenting and classifying tooth types on the 3D dental models. First, the tooth classification method capitalized on the two-level hierarchical feature learning is proposed to solve the misclassification problem in highly similar tooth categories. Second, we exploit an improved three-level hierarchical segmentation method based on the deep convolution features to conduct segmentation of teeth-gingiva and inter-teeth, respectively, and the conditional random field model is used to refine the boundary of the gingival margin and the inter-teeth fusion region. The experimental results show that the classification accuracy in Level_1 network is 95.96%, the average classification accuracy in Level_2 network is 88.06%, and the accuracy of tooth segmentation is 89.81%. Compared with the existing state-of-the-art methods, the proposed method has higher accuracy and universality, and it has great application potential in the computer-assisted orthodontic treatment diagnosis.

/pdf/automatic-classification-and-segmentation-of-teeth-on-3d-2g0kzbqd6c.pdf

Automatic Classification and Segmentation of Teeth on 3D Dental Model Using Hierarchical Deep Learning Networks

TSegNet: An efficient and accurate tooth segmentation network on 3D dental model

Interactive tooth partition of dental mesh base on tooth-target harmonic field

Autonomous vehicles (AVs) must share space with pedestrians, both in carriageway cases such as cars at pedestrian crossings and off-carriageway cases such as delivery vehicles navigating through crowds on pedestrianized high-streets. Unlike static obstacles, pedestrians are active agents with complex, interactive motions. Planning AV actions in the presence of pedestrians thus requires modelling of their probable future behavior as well as detecting and tracking them. This narrative review article is Part I of a pair, together surveying the current technology stack involved in this process, organising recent research into a hierarchical taxonomy ranging from low-level image detection to high-level psychology models, from the perspective of an AV designer. This self-contained Part I covers the lower levels of this stack, from sensing, through detection and recognition, up to tracking of pedestrians. Technologies at these levels are found to be mature and available as foundations for use in high-level systems, such as behavior modelling, prediction and interaction control.

/pdf/pedestrian-models-for-autonomous-driving-part-i-low-level-35grjcf04k.pdf

Pedestrian Models for Autonomous Driving Part I: Low-Level Models, From Sensing to Tracking

Current research on skeleton tracking techniques focus on image processing in conjunction with a video camera constrained by bones and joint movement detection limits. The paper proposed 3D skeleton tracking technique using a depth camera known as a Kinect sensor with the ability to approximate human poses to be captured, reconstructed and displayed 3D skeleton in the virtual scene using OPENNI, NITE Primesense and CHAI3D open source libraries. The technique could perform the bone joint movement detections in real time with correct position tracking and display a 3D skeleton in a virtual environment with abilities to control 3D character movements for the future research.

Skeleton Tracking using Kinect Sensor & Displaying in 3D Virtual Scene

This paper presents the computerized segmentation algorithm on the 3D digital dental model for separating individual tooth using Borland C++ Builder® and OpenGL The purpose is to apply the algorithm in the 3D orthodontic diagnosis and treatment software package The method consists of 5 main steps as follows First the digital dental model is produced by 3D scanner in STL format and the occlusal plane is calculated Second the arch form of the model is approximated Third the panoramic of the dental cast in 3D system is constructed Fourth, the 3D-panoramic is converted to 2D system to find the cutting points for segmentation of individual tooth, and finally the 2D image is converted back to 3D system for creating the vertical plane for separating teeth into difference colors at the cutting points The result has shown that the 3D digital dental model can be segmented for separating individual tooth, and the algorithm is ready to apply in orthodontic diagnosis and treatment software package

http://www.cephsmilev2.com/chanjira/Papers/2010/eng/060-V1-E165.pdf

Computerized algorithm for 3D teeth segmentation

The method of three dimensional mesh reconstruction is used to supplementing the 3D reconstruction or 3D scanning method, which only provide with point cloud in 3D model However, the 3D models in mesh mode illustrate clearer appearance and can be used to obtain more information In addition, the 3D reconstruction has been previously developed by employing the digital fringe pattern projection where the output was formed in the point cloud Therefore, this paper presents the computerized algorithms to construct the 3D mesh from the point cloud The algorithm can be done by firstly dividing the point cloud into small volume for parallel calculation Next, the points in each small volume are connected, which becomes the submesh and then integrated them together The submesh in each small volume is produced by creating the initial triangle and then expanding triangular construction continually from each side (base) which has been previously generated The expanded triangles must be satisfied some certain constrains The point cloud which is filtered from the STL file format was tested using proposed method The mesh reconstruction is shown with reasonable result

Method of 3D mesh reconstruction from point cloud using elementary vector and geometry analysis

Interactive Virtual 3D Gallery Using Motion Detection of Mobile Device

For awareness of becoming parents, this paper proposes the development of Z-Baby as an android application to calculate the pregnancy due date, fetus development simulation, weight gain during pregnancy and the amount of calorie in needs. Pregnancy due date is calculated based on user defined data which is the last menstruation. Next, user can view the simulation of 3D fetus development with description week by week using OpenGL ES. Then the weight gain during pregnancy is calculated from the BMI based on the specification by IOM 2009 while allowing a user to manage her weight gain as well. Updated weight data were plotted dynamically using Android Plot API. Daily calorie requirement can be calculated during pregnancy by Harris-Benedict Equation to maintain the appropriated weight for healthy baby without overeating. Other available functions are recording photos & data in monthly basis, locating mom's position on Google map to find the nearest hospital located on the map. For accuracy testing, Z-Baby has been compared to other software packages which have the same functions. Z-Baby in now available for free in Google Play. Keywords—Fetus development; Weight gain during pregnancy; Calorie requirements; OpenGL ES; Android plot API.

Nonlapas Wongwaen

Papers

Skeleton Tracking using Kinect Sensor & Displaying in 3D Virtual Scene

Computerized algorithm for 3D teeth segmentation

Method of 3D mesh reconstruction from point cloud using elementary vector and geometry analysis

Interactive Virtual 3D Gallery Using Motion Detection of Mobile Device

ZBaby: Android application for pregnancy due date, fetus development simulation and weight gain during pregnancy