Xiuli Ma

Texture image segmentation on improved watershed and multiway spectral clustering

Spectral clustering is a new graph and similarity based clustering algorithm. When the image is too big, it will take a long time to compute affinity matrix and its eigenvalues and eigenvectors. In order to improve the convergent speed of spectral clustering, a two-stage texture segmentation algorithm is proposed in this paper. First, an improved watershed algorithm is used to perform pre-segmentation and then multiway spectral clustering with eigenvalue-scaled eigenvectors performs the final segmentation. This can reduce the runtime greatly and it is valuable to application with high time request. To verify the proposed algorithm, it is applied to texture image segmentation and the segmentation results are satisfying.

An adaptive threshold algorithm based on wavelet in QRS detection

Electrocardiogram (ECG) signal has been widely used for cardiac diagnostic and pathological analysis. The QRS complex is the most string waveform within the ECG, and it carries large amounts of information. But it is always mixed with various noises, such as power line interference, baseline displacement and electromyography (EMG) noise. These noises bring obstacle to the diagnosis of cardiovascular diseases. Hence, eliminating these noises is the necessary and important way to analyze ECG signal. In this paper we elaborate a new algorithm based on wavelet transform and adaptive threshold after hundreds of experiments and simulations using MIT-BIH arrhythmia database signals. According to the results, this algorithm can remove noises efficiently with low distortion, so it can also meet the needs of the clinical treatment and pathological research.

Accelerated algorithm for 3D intestine Volume Reconstruction base on VTK

3D Volume Reconstruction of the intestine is a crying need to improve the accuracy of disease detection and treatment. This paper adopts the Visualization Toolkit (VTK) combined with VC++ IDE to reconstruct 3D intestine CT images. Ray-Casting is a traditional algorithm for 3D Volume reconstruction of medical images, but the calculation is very large and the speed of creating images is so slow-footed. To overcome the drawback of large quantity of computation, firstly, in order to effectively reduce the amount of redundant data of reconstruction, segment the intestinal part which we interest in by using Region Growing method, then, at the step of image synthesis, set an appropriate threshold of opacity, when the opacity is greater than the threshold value, that means the sampling points behind will not contribute to the color value of current pixel, so terminate the calculation of synthesis, and finally, quickly rebuild 3D intestine. Compared with the traditional algorithm, the accelerated algorithm achieves great enhancement in efficiency, and without significantly reducing image quality.

Cloud model-based control strategy on cluster communication coverage for wireless sensor networks

For random deployment in sensor networks, there exists uneven node density, seriously reducing the network performance after clustering. The idea of the power control in CDMA is applied to handle the communication coverage control for clusters in wireless sensor networks. The cloud model-based uncertainty reasoning and control mechanism are introduced to adjust the transmit power adaptively to keep the node number within each cluster in an appropriate range in accordance with the node distribution density. The tendentious rules of cloud model ensure the convergence performance of the coverage control, while random process concealled in cloud model ensures the best iteration times in all. After clustering, there exists an appropriate node number within each cluster, improving the network topology. The experiment results validate its rationality and effectiveness.

3D visualization for heart model from point clouds

In order to improve the efficiency of the ablation operation, this paper proposes a 3D visualization system. In the system, cloud section is used, looking for outer points and eliminating noise-points effectively. And then we adopt Poisson surface reconstruction to achieve 3D reconstruction for scattered point-cloud, which can obtain more ideal geometry. Then, a further optimization is made for the model. The test shows that this system has good reliability and instantaneity.

Research and implementation of shortest path algorithm on PLY triangular mesh model

Three-dimensional virtual surgery has received a significant amount of attention in recent years. It is necessary to mark organ lesions and add ablation path on virtual organ model. The contribution of this paper is that we propose and implement a new shortest path algorithm based on PLY triangular mesh model. The algorithm follows the topology structure of model and is simple to implement. We add ablation line on three dimensional heart model efficiently and correctly utilizing the proposed algorithm and have a comparison with some traditional algorithms. The experiments show that the proposed shortest path algorithm is not likely to deviate from the target node compared to Dijkstra, moreover, the speed of our algorithm meets the needs of the project very well.

Traffic sign recognition based on kernel sparse representation

This paper proposes a novel approach based on scale invariant feature transform (SIFT) and kernel sparse representation for traffic sign recognition in complex traffic scenes. This module consists of several steps. In the first stage, SIFT is introduced for feature extraction from samples and test targets, respectively. The features are mapping to the kernel space. In the second stage, we construct an over-complete dictionary based on kernel sparse representation. Finally, traffic objects are recognized by computing sparseness and reconstruction residuals in the dictionary. Experiment results show that the proposed approach enhances the class discriminant ability using traffic features with higher recognition preciseness and robustness in complex traffic scenes compared with SVM, SRC.

Study on simulation of catheter-heart interaction based on BD-tree

Medical catheter plays a very important role in the minimally invasive cardiac surgery. Doctors want to know the real-time location of a catheter when the catheter is navigating in the heart. This paper studies on simulation of interaction between heart and catheter based on BD tree. The heart data are collected from hospital and a heart model is reconstructed using the Poisson algorithm. The catheter tip is simulated by a sphere, and its movement is controlled by keyboard. The nearest point of the catheter tip on the heart model is got in real time based on BD tree. The distance between the catheter tip and the heart is calculated by the distance between the center of sphere and the nearest neighbor point in the heart model. A projection of the catheter tip in the heart model is applied, while the color of the projection changes along with the distance. The development platform is OpenGL combined C++ language.

Local subdivision on triangle mesh

Subdivision has been a very popular technology in graphics at home and abroad. It can be splited into global subdivision and local subdivision, and the methods to subdivide are various including butterfly, modified-butterfly, loop, linear, Catmull-Clark, Doo-Sabin and so on. These methods can be divided into interpolation subdivision and approach subdivision. In this paper, it mainly involves approach and interpolation algorithm, and it mainly introduces a local subdivision algorithm based on triangle mesh reconstruction by Poisson. Aiming at the global subdivisions disadvantages, it is not only save memory space and shorten run time. Local subdivision is very popular in various fields In many case, global subdivision is too waste resource and low efficiency, so global subdivision is not a fine method to optimize model. According to the request and experimental verification, linear algorithm is more suitable for the subject in this paper.

An improved subdivision algorithm using for heart model

In this paper, we use a three-dimensional heart model which is reconstructed by cloud points. As the need of post-processing such as adding ablation line and opening on the model, we have to subdivide the grid model. After analyzing the characters of the model, basing on the loop algorithm, we propose a method in which we subdivide a triangle into 16 triangles, and through the experiment we get a very uniform sub-grid model and a much more smoothly 3D heart surface model. In this paper we also make a contrast by opening on the original model and the subdivided model, and we find the cross-section on subdivided model is much more smoothly. The speed of the postprocessing on the subdivided model also meets the needs of the project very well.