Xukun Shen

A SPH-based method for interactive fluids simulation on the multi-GPU

In this paper, we present a Smoothed Particles Hydrodynamics(SPH) implementation algorithm on Multi-GPU which is used for physics-based interactive fluid animations by the parallel framework. We employ the SPH method of a particle-based pure Lagrangian approach to discretize Navier-Stockes equation for diverse fluid animations. In order to simulate the incompressibility of liquid to the utmost extent while assuring numerical stability of the system, we use a improved Tait equation to compute pressure. For low computational expense of each simulation step, combining the characteristics between the CPU and GPU, we introduce index sort neighborhood search method which uses CUDA architecture and eliminates GPU memory overhead and saves searching time. In order to get some vivid and interactive fluid effects, we apply an image spaced method to capture the refractive effect and an adaptive method to generate the caustic map for each light. The implementation has been highly optimized to the point where a scaled simulation can run in real-time with CUDA. On the Multi-GPU platform, we obtain good acceleration and high quality rendering effect. In the conclusion, we demonstrate the quality and performance of our method for animating different scale and scene fluid interactive experiments.

The augmented reality research progress in collaboration environment of CAR-CA

It is a new aspect of engineering science that uses augmented reality and collaboration technology to aid the design and maintenance of large equipments. To meet the requirement of the design and manufacture of commercial aircraft in China, CAR-CA (Collaborative Augmented Reality for Commercial Aircraft Design and Maintenance) is developed now to aid the cockpit design and training of components disassembly and assembly. This paper introduces the engineering and technology progress of CAR-CA, including the system architecture, system components, technology researches and future work.

Image-Based Hierarchical Method of 3D Model Retrieval

In this paper, we present a novel and hierarchical image-based retrieval method of 3D shape by using categorization tree. Unlike other methods, we can get the accurate matched 3D shape by comparing only a small subset of the model database. The central idea of our method is that the categorization tree is used to organize all 3D shapes and form a representative subset of the database. The retrieval is divided into two stages: we first get a temp matched shape from the subset, and then obtain the final matched shape from the shapes which are close to the temp matched shape in the categorization tree. In the process of search, we use the Suggestive Contour to improve the accuracy of the matching. To speed up the search, we adopt the min-hash and the inverted index techniques. Experiments on database demonstrate accuracy, robustness and efficiency of our method.

A Face Matching Method Based on Biometric for Surgical Navigation System

Fast, accurate registration between image space and patient space is the key point to surgical navigation system. In this paper, we research and implement an face matching method based on biometri. Against to the smooth, textured feature of human face, we built a proactive structured light vision system, to get patients facial point cloud, proposed and implemented a rough face point cloud registration method based on biometric feature of human face. By identify the nose and eyes in the two-dimensional image space by Active Shape Model (ASM) method which based on statistical information, and identify the corresponding nose and eye by the principal component analysis of the local neighborhood in the three-dimensional space, thereby automatically get an initial registration matrix for those two point cloud. Then accurate complete point cloud registration, based on Iterative Closest Point (ICP) algorithm, achieving space registration.

Automatic Image Annotation with Real World Noisy Data

Automatic Image annotation is an important open problem in computer vision. In real world dataset environment, image labels are often noisy. For the task of image annotation with weakly labels, we propose SNLWL, a semantic neighborhood learning model from weakly labeled dataset. Missing labels are replenished using reweighting the error loss function. Then semantic balanced neighborhood is construct for samples in the training set. The methods allows the integration of multiple label metric learning and local nonnegative sparse coding. In this manner, we can optimally construct semantic consistent neighborhood where neighbors have higher global similarity, partial correlation, conceptual similarity along with semantic balance for samples in the training set. We also introduce an iterative denoising method of the label predictions to handle the noise. We investigate the performance of different variants of our method and compare to existing work. We present experimental results for various data sets. On all datasets, SNLWL makes a marked improvement as compared to the current state-of-the-art.

Research on SPH Parallel Acceleration Strategies for Multi-GPU Platform

This paper proposes an acceleration strategy for SPH on single-node multi-GPU platform. First the acceleration strategy for SPH on single-GPU is studied in conjunction with the characteristics of architecture. Then the changing pattern of SPHs computation time has been discussed. Based on the fact that the changing pattern is rather slow, using a simple dynamic load balancing algorithm an acceptable load balance is obtained on multi-GPU. Finally, an almost linear speedup is achieved on multi-GPU by further optimizing dynamic load balancing algorithm and communication strategy among multiple GPUs

GPU-BASED FLUID SIMULATION WITH FAST COLLISION DETECTION ON BOUNDARIES

In this paper, we present a method for fluid simulation based on smoothed particle hydrodynamic (SPH) with fast collision detection on boundaries on GPU. The major goal of our algorithm is to get a fast SPH simulation and rendering on GPU. Additionally, our algorithm has the following three features: At first, to make the SPH method GPU-friendly, we introduce a spatial hash method for neighbor search. After sorting the particles based on their grid index, neighbor search can be done quickly on GPU. Second, we propose a fast particle-boundary collision detection method. By precomputing the distance field of scene boundaries, collision detections computing cost arrived as O(n), which is much faster than the traditional way. Third, we propose a pipeline with fine-detail surface reconstruction, and progressive photon mapping working on GPU. We experiment our algorithm on different situations and particle numbers of scenes, and find out that our method gets good results. Our experimental data shows that we can simulate 100K particles, and up to 1000K particles scene at a rate of approximately 2 times per second.

SPH-Based Fluid Simulation: A Survey

This paper presents an up-to-date survey on the application of Smoothed Particle Hydrodynamics (SPH) method in fluid simulation. As a pure Lagrangian method, it has a wide range of applications in computer graphics. We summarize it from key issues of kernel function, neighborhood search, boundary conditions, time integral to algorithm application in diverse fluids, including bubble, fluid-solid interaction, multiphase flow and mixed flow. We also discuss the hotspots area, including adaptive sampling, flow control, surface rendering and accelerating algorithm. In experiment section, we show four different animation scenes based on the survey algorithm in our platform. In the end, we point out the potential problems and prospects for future work.