Recently,the Support Vector Machine(SVM) using Spatial Pyramid Matching(SPM) kernel has achieved remarkable successful in image classification. The classification accuracy can be improved further when combining the sp...
详细信息
Recently,the Support Vector Machine(SVM) using Spatial Pyramid Matching(SPM) kernel has achieved remarkable successful in image classification. The classification accuracy can be improved further when combining the sparse coding with SPM. However,the existing methods give the same weight of patches of SPM at different levels. Clearly the discriminative powers of SPM at different levels are distinct and there are correlation relationships among the sparse coding bases vectors,which usually have negative influence on the classification accuracy. This paper assigns different weights to the patches at different levels of SPM,and then proposes a new spatial pyramid matching kernel. Furthermore,the Principle Component Analysis(PCA) is employed to reduce the dimension of the feature vectors in order to decrease correlation among vectors and speed up the SVM training process. The preprocessing can enhance the discriminative ability of the new kernel as well. Experiments carried out on Caltech101 and Caltech256 datasets show that the new SPM kernel outperforms the existing methods in terms of the classification accuracy.
We present a physically-based animation method for simulating the diffusion of water pollutant in this paper. Our method allows animating the dynamic evolution of biological pollutants(e.g., water hyacinth, algae) on ...
详细信息
We present a physically-based animation method for simulating the diffusion of water pollutant in this paper. Our method allows animating the dynamic evolution of biological pollutants(e.g., water hyacinth, algae) on largescale water surface effectively. To enable the simulation of such a phenomenon in large scale virtual environment, we simplify the problem by considering the water pollutants as large patches or clusters of aquatic plants rather than each single aquatic plant. Since there exits obvious interface between the large patches of pollutants and the water surface, we model the evolution of the interface in our approach. We use 2D dynamic curves to represent the interface and the diffusion of the water pollutants can be regarded as the dynamic evolution and propagation of 2D curves. To alleviate the topological changes of 2D curves, we adopt a physically-based 2D level set model to animate the evolution and propagation of the interface. We build a level set equation to model the evolution of the interface. In addition, to handle the large scale virtual environment correctly in our physically-based level set model, an imagebased 2D voxelization method is proposed in the paper. In the voxelization method, the virtual environment will be converted to boundary conditions when solving the level set equation. Finally, the water pollutants diffusion phenomenon is simulated on large scale water surface by merging the interface animation results as well as the large scale virtual environment. Animation results about the algae propagation phenomenon in Taihu Lake show that our method is intuitively to be implemented and very convenient to produce visually interesting results.
computer vision-based fire detection involves flame detection and smoke detection. This paper proposes a new flame detection algorithm, which is based on a saliency detection technique with prior information in color ...
详细信息
ISBN:
(纸本)9781479944187
computer vision-based fire detection involves flame detection and smoke detection. This paper proposes a new flame detection algorithm, which is based on a saliency detection technique with prior information in color space. In still images and video sequences, an area containing a flame always attracts the attention because it is an exceptional event. Thus, to utilize the color information of flame pixels, the probability density function of the flame pixel color can be obtained using Parzen window nonparametric estimation. This prior probability density function is then fused with the saliency detection phase as topdown information so the flame candidate area can be extracted. According to the experimental results, our proposed method can reduce the number of false alarms greatly compared with an alternative algorithm, while it also ensures the accurate classification of positive samples. The classification performance of our proposed method was better than that of alternative algorithms.
In crowd animation production, it is difficult to design realistic and natural crowd motion paths using traditional methods. In order to solve the problems in generating paths of a crowd, a new approach based on an in...
详细信息
In crowd animation production, it is difficult to design realistic and natural crowd motion paths using traditional methods. In order to solve the problems in generating paths of a crowd, a new approach based on an interactive evolutionary algorithm(EA) named biogeography-based optimization(BBO) is proposed in this paper. The optimization process of individuals is considered as the motion process of a crowd. Specifically, the new approach is applied into a system to generate crowd motion paths and the simulation results demonstrate the efficiency and effectiveness of it.
Community detection is a long-standing yet very difficult task in social network analysis. It becomes more challenging as many online social networking sites are evolving into super-large scales. Numerous methods have...
详细信息
Proper understanding of textual data requires the exploitation and integration of unstructured and heterogeneous scientific literature, which are fundamental aspects in literature retrieval research. The traditional l...
详细信息
Proper understanding of textual data requires the exploitation and integration of unstructured and heterogeneous scientific literature, which are fundamental aspects in literature retrieval research. The traditional literature retrieval is based on keyword matching, and the retrieval results often deviating from the users' needs. In this paper from the perspective of ontology, we built shareable and relatively perfect medical enzyme ontology, which is the foundation of the study of domain ontology constructing method. The ontology-based full text retrieval algorithm is put forward, and a document retrieval system based on medical enzyme semantics is designed and implemented, which can not only implement intelligent literature retrieval, but also improve the recall significantly while keeping high precision. This system can employ in particular area moreover it can be used in different areas of the semantic retrieval, which can provide intelligent foundation for the expert systems in medical enzymes field, information retrieval and natural language understanding, etc. The experimental results on the public medical enzyme domain dataset show that our approach performs better than the state-of-the-art methods.
The article puts forward a new formula of computing the popularity of the topics, which is based on the stream of the news on the internet. What's more, the features of the microblogs should be also considered and...
详细信息
In mobile underwater environment, underwater wireless sensor networks (UWSNs) keep moving and dispersing due to water flowing and aquatic creatures touching, and thus some isolated nodes appear. This type of isolated ...
详细信息
In mobile underwater environment, underwater wireless sensor networks (UWSNs) keep moving and dispersing due to water flowing and aquatic creatures touching, and thus some isolated nodes appear. This type of isolated nodes cannot obtain enough anchor nodes in their communication range, which makes self-localizations disabled. In order to solve this problem, a multi-hop localization scheme is proposed in this paper. Firstly, ordinary nodes between anchors and unknown nodes are set as routers to find the shortest paths by a greedy approach; secondly, the shortest paths are approximately fitted into a straight distance between two nodes; finally, the positions of unknown nodes can be calculated by trilateration. The proposed algorithm is simulated and is compared with other algorithms in terms of localization error, and the results are proven preferable.
To balancing energy consumption in wireless sensor networks, we proposed a fixed time interval node broadcasting scheme under variational acceleration straight-line movement model. Simulation results show that the app...
详细信息
To balancing energy consumption in wireless sensor networks, we proposed a fixed time interval node broadcasting scheme under variational acceleration straight-line movement model. Simulation results show that the approach proposed in this paper has a superior performance on energy consumption balance compared to uniform broadcasting methods.
The Natomas basin is a 21,450 hectare (53,000 acre) area immediately north of downtown Sacramento. The area, which includes an international airport, a civic sports arena, commercial developments, and extensive reside...
详细信息
暂无评论