Papers by Author: Xue Jun Xu

Abstract: Interactive segmentation with graph cuts has become very popular and many priors have been introduced into graph cuts to improve the results. This paper proposed a method which uses the deformable part-based model to pre-label the seeds. First the deformable part-based model finds out the bounding box, then we can pre-label the seed point based on the assumption of compact shape. Our results show that our method can get more accurate result especially the appearance of the object and background are similar and the shape is compact.

Abstract: To give a relative accurate detection result in the computer vision which apply in the Video Surveillance and so on. A method based on the auto collects the seeds and then use the Cellular atuomata to subtract the moving object. Firstly we use the frame difference image to find the moving region and give the seeds of the foreground and background. Then we use the grow cut of CA to cut the frame into the foreground and background. The experiment is shown our seeds can give more accurate information of the foreground and get a relative precise result.

Abstract: Matching feature points is an important step in image registration. For high- dimensional feature vector, the process of matching is very time-consuming, especially matching the vast amount of points. In the premise of ensuring the registration, filtering the candidate vectors to reduce the number of feature vectors, can effectively reduce the time matching the vectors. This paper presents a matching algorithm based on filtering the feature points on their characteristics of the corner feature. The matching method can effectively improve the matching speed, and can guarantee registration accuracy as well.

Abstract: Image processing toolbox of Matlab provides a morphological skeletonization operator. But this function is sensitive to noises. The aliasing shape's skeleton computed by this operator involves many extra branches. In order to eliminate the influence of aliasing noises, a combined method is introduced. The method combines the morphological thinning and skeletonization operators, and takes their advantages. It is not sensitive to noises, and can eliminate the extra branches.

Abstract: Geographic Information Systems (GIS) are designed to support the management, manipulation, analysis, and modeling of spatial data. Water supply pipeline network is a system which has a large of spatial attribute data. In order to improve the managing and working efficiency, the paper proposed a new method for designing the long distance pipeline based on GIS. Seamless integration of GIS platform, hydraulic model of long distance pipeline is built which is base of hydraulic calculation. The model is based on building pipeline concise model dynamic. The hydraulic calculation is to guarantee the authenticity of the model which can help to simulate the whole system status and analyze the condition of long distance pipeline. The pressure and safety coefficient models of long pipeline based on GIS can help operator to know the details of every segment in the long distance pipeline system. The method locates the weakness of long distance pipeline and displayed impressively on GIS platform. All the technologies strengthen the security of the pipeline design, improve the accuracy of hydraulic calculation, and help operators to analyze the long distance pipeline condition by spatial visualization. The example shows that the method is reasonable, effective, and has been applied to practical project.