Applied Mechanics and Materials Vol. 311

Abstract: Determining optimum cluster number is a key research topic included in cluster validity. Based on geometric probability, this article proposes a new cluster validity algorithm to determine optimum cluster number for two-dimension datasets. The algorithm measure the cluster structure of the data set according to the distributive feature of the point set in the characteristic space. The structure information of the point set has been stored in a line segment set generated by connecting each pair points in the point set and the cluster validity function is formed by comparing the values of line segment direction in the line segment set with those resulted from completely random condition. Experiments prove that the pattern of the function curve generated with a given example data set effectively enables determining the optimum cluster number of the data set.

Abstract: The on-line Class Constrained Bin Packing problem (CCBP) is one of variant version of the Bin Packing Problem (BPP). The BPP is to find the minimum numbers of bins needed to pack a given set of items of known sizes so that they do not exceed the capacity B of each bin. In the CCBP, we are given bins of capacity B with C compartments and n items of Q different classes, each item i is belong to 1,2,…,n with class _qi and _si. The CCBP is to pack the items into bins, where each bin contains at most Q different classes and has total items size at most B. This CCBP is known to be NP-hard combinatorial optimization problems. In this paper, we used an ant colony optimization (ACO) approach with a simple but very effective local search algorithm to resolve this NP-hard problem. After the experimental design, limited computational results show the efficiency of this scheme. It is also shown that the ACO approach can outperform some existing methods, whereas the hybrid approach can compete with the known solution methods.

Abstract: We propose a novel adaptive fast learning (AFL) algorithm for two-dimensional principal component analysis (2DPCA) in this paper. As opposite to conventional PCA which is based on 1D data vectors, 2DPCA is based on 2D image matrices and thus has higher accuracy than conventional PCA when applied to applications such as face recognition, facial expression recognition, palmprint recognition, etc. Our proposed AFL algorithm simultaneously estimates both eigenvectors and corresponding eigenvalues, and then adaptively sets the learning rate parameters of neurons to ensure all neurons learning with almost the same fast speed. Requiring no image covariance matrix evaluation, the desired multiple eigenvectors of 2DPCA can thus be learned effectively in the form of weight vectors of neurons. The proposed AFL algorithm can also be applied to learning for T-2DPCA. Simulation experiments performed on face database such as the YaleB database clearly demonstrate that the proposed AFL algorithm performs very well and thus is a very effective computational tool for both 2DPCA and T-2DPCA.

Abstract: In this paper, we propose a new approach to accelerate the pocketing tool-path generation by using graphic hardware (graphic processing units, GPU). The intersections among tool-path elements can be eliminated with higher efficiency from GPU-based Voronoi diagrams. According to our experimental results, the GPU-based computation speed was seven to eight times faster than that of CPU-based computation. In addition, the difference of tool-path geometry between the CPU-based and GPU-based methods was insignificant. Therefore, the GPU-method can be efficiently used to accelerate the computation while the precision is assured for the tool-path generation in pocketing machining.

Abstract: Facebook is currently one of the world's most popular social networking services, and has been widely used in the field of e-learning. In general, learners in e-learning environments need to evaluate their learning ability through taking tests to present the learning achievement. In order to evaluate their ability on e-learning platform with social network services, this study proposes an automatic question generation system for individual learning status. The proposed system uses the artificial bee colony algorithm to find suitable questions for each learner according to the learner's profile, reading experience, professional ability, and the e-learning records in the system. The experimental results indicate that the proposed method improves the accuracy of the automatic question generation system and that it outperforms the random method.

Abstract: 3D walkthrough (henceforth referred as WT) becomes popular and apparent when the volume of query processing in some 3D scenery (e.g., the walkthrough system and the 3D museum navigation systems) is considered. In this scenario, different people with similar behaviours induce completely different space-time traversal patterns in a recoded traversal sequence. This is because they navigate different paths, and their surrounding backgrounds are different. What are common across such sequences of the same behaviours are the underlying induced walkthrough fields. We want to make use of regions of common sequential traversal patterns for acceleration and compaction purposes. In this paper we propose a new Sequence-based Pattern Similarity (SPS) approach based on a behavioural walkthrough system that exploits sequence-based semantic-oriented clustering techniques, such as association, intra-relationships, and inter-relationships, to explore additional links throughout the behavioural walkthrough system. The experimentation shows that in such cases the use of so-called I/O-efficient algorithms that minimize the number of disk accesses can lead to tremendous runtime improvements.

Abstract: Online rapid three-dimensional reconstruction is widely applied in virtual reality, heritage preservation, bio-engineering and architectural fields. The error caused by image quality or manual import is the main reason for the low quality of model details when applying current reconstruction methods while meeting the time premise. To solve this problem, the paper proposes a fast and smooth carving algorithm for online 3d reconstruction by joining the filter. By applying the method, you can get a more realistic and smooth three-dimensional reconstruction results. First, we convert the input point cloud to meshes through Delaunay tetrahedralisation. Then we reconstruct the model with the space carving algorithm with the filter to obtain the result. The experiment result shows our method exceeds existing methods while meeting the time constraints under the premise at the same time.

Abstract: MapReduce framework of cloud computing has an effective way to achieve massive text categorization. In this paper a distributed parallel text training algorithm in cloud computing environment based on multi-class Support Vector Machines(SVM) is designed. In cloud computing environment Map tasks realize distributing various types of samples and Reduce tasks realize the specific SVM training. Experimental results show that the execution time of text training decreases with the number of Reduce tasks increasing. Also a parallel text classifying based on cloud computing is designed and implemented, which classify the unknown type texts. Experimental results show that the speed of text classifying increases with the number of Map tasks increasing.

Abstract: This paper present a family safety monitoring system, which integrated function of telephone alarm, remote control by telephone network and Ethernet. Host alarm system, controlled by MCU, collects data wirelessly from sensors, and makes remote alarm by DTMF. The video processing and transmitting system, with a DSP chip of TMS320F28335 as the core processor, transmits the video data by means of Bluetooth indoors, and by Ethernet in the residential district, so that remote control through Internet can be implemented.

Abstract: Biometric is used to confirm the unique of identity. In general, face is the most characteristic to recognize a person. In this paper, it is emphasized and compared the quality of 2D and 3D face recognition. There are three parts in this paper. First part is the detection of skin color which is used RGB color space. In order to reduce color red and green which are sensitive to illuminant, Normalized Color Coordinate (NCC) method is chosen to pick up the range of skin color directly. Second, to increase choosing of the important characteristics by Principle Component Analysis (PCA) the wavelength distinguishes technique is used to make 3D images. The third part is about identifying. An improved PCA through a transfer matrix to get optimal total scatter matrix of within-class scatter matrix is used. The optimal total scatter matrix represents the eigenvalue of face characteristics. Finally, the recognition rate and process performance between 2D and 3D images are compared via Euclidean Distance. The efficiency and recognition rate of 3D images are superior to 2D images. The recognition rate of 3D images attains to 92% and costs 0.39 second to recognize each image. It is improved 28% compared with the recognition rate of 2D images.