Due to the limited length and freely constructed sentence structures, it is a difficult classification task for short text classification, especially in multi-class classification. An efficient meta learning framework is proposed for twitter classification. The tweets are clustered into many sentence styles corresponding to new class labels. Thus, the original text classification task becomes few-shot learning task. When applying few-shot learning on benchmark datasets, the proposed method Meta-CNN achieves improvement in accuracy and F1 scores on multi-class twitter classification, and outweigh some traditional machine learning methods and a few deep learning approaches.

Because the micro lenses segment the plenoptic image physically, the global optimization method is hard to be applied on the depth estimation progress and the WTA (winner-take-all) method based on the multibaseline system will generate very coarse but continuous depth map. The global optimizations method always generate robust but concrete depth map when using multi-label technique. So cost projection strategy by directly processing the lenslet plenoptic image is proposed to build the cost volume based on the projection image for global optimization. A method based on Markov Random Fields Propagation (MRFsP) is adopted by merging the modified WTA depth result to refine the concrete depth map. For validation, the depth map built with the proposed method is compared with the microlens-based optimization result, and the result shows obvious improvement.

In order to improve the pixel distinguishability, this paper introduces an algorithm that the quantized geometric relationship among sub images of light field replaced the pixel value. Based on this relationship, the epipolar line division ratio is calculated to take place of pixel value. Experimental results showed that the disparity map with the division ratio images are obviously superior to those obtained by using the subaperture images.

Using the matrix expression form of computer vision projection equation, the collinear equation is constructed into matrix equation. With the projection matrix element as a composite function, this paper realizes the unification derivation of each variable of the collinear equation based on the matrix analysis method. Compared with the traditional analytical method of linearization, the form of matrix analysis process is quite succinct and easy to understand, which can be used to the numerical solution of linear library application. For the various construction form of the rotation matrix, this method has better adaptability. The constructed matrix of collinear equation has important enlightenment significance for using computer vision method.

The authors explore an approach on large scale of regional ecological risk assessment, and implement the integrated assessment of China, so as to provide scientific support for eco-system risk recognition and risk prevention. Select the 10 natural disasters as ecological risk sources, 22 species of the ecosystem as risk receptors and consider eco-environmental vulnerability. The integrated ecological risk assessment and mapping are conducted, on the basis of evaluating ecological risk by individual disaster and overlaying various risk evaluation by GIS. The result shows that the high, medium risk level of eco-system areas account for 45% of China’s territory. The environmental protection and risk prevention are very important and need to go a long way. It is imperative to establish adaptive disaster reduction and risk prevention mechanisms, make rational use and conservation for the forests, grassland, farmland, wetlands and other ecosystems.

An orienting method is proposed for bionic navigation with the polarized light patterns. Combining theoretical analysis with experiment, authors build a kind of theoretic and experimental models. Then the sun’s relative position and a navigation reference line (solar meridian) are derived details for navigation system to provide orientation information. Finally, based on the horizon coordinate system, an example is given for the whole calculate process of direction angle.

This paper introduces the wavelet tool and 4 indexes to quantify the temporal characteristics of tourist flows. Based on the monthly visitor data from 1994-2006 and daily visitor data from 2003-2006 which were mainly collected through JDS (Jiuzhaigou digital system), the multi-time scales characteristics of tourist flow along the time series were analyzed by using the morlet wavelet transform. The complex structure of multi-time scales are exhibited, also, the periodic oscillation and the main periods of tourist flows at different scales are identified through Wavelet Toolbox in MATALAB 7.3. The results show that 1) during 1994-2006, the number of tourists increased rapidly in the way of cubic function and curve of tourists had apparently 3 peaks in July-August, October, and May respectively; 2) at the month scale, the trend of tourist fluctuation decreased overtime; 3) in the past 13 years, there was an approximately 1 year period in the tourist time series. Furthermore, 50-70 days was another period in 2003-2006. After applying wavelet tool to analysis tourist flows, this article argues that the localization characteristics of time-frequency for wavelet analysis can effectively demonstrate the detailed structures of tourist flows. Wavelet is a promising method to analyze temporal characteristics of tourist flows, and the findings can provide a scientific guideline for adjusting and controlling tourist flows.

A digital bathymetry model according to principle of multi-beam bathymetric sonar system is designed, and a correction method of water depth survey due to effect of angular beamwidth is given. Accordingly, terrain maps are constructed to mimic the realtime terrain. The model and correction are verified by using practical water depth data of certain seafloor. Simulation results show that the model is in accordance with theoretical analysis and practical situation, and the correction effect of water depth is obvious.

Seabed terrain matching is applied to increase the localization accuracy for autonomous underwater vehicles (AUV). By bicubic B-spline surface interpolation, reference depth map (RDM) is reconstructed to represent actual terrain. A plot is taken from RDM plus white noise to mimic local depth map (LDM). Then two maps are transformed to 8-bit,256-grey images. In this stage, the two images are matched with correlation, thus image and terrain are connected tightly. Simulation results show that the matching is greatly accurate with certain robusticity, suitable to large seabed terrain areas, including flat and complex regions.

An observator is combined with fuzzy set theory and neural networks for monitoring transformer operation-state. It has overcome the limitations such as the dependency on experts for fuzzy rule and no self-adaptation for fuzzy set. The self-regulation of membership functions and the self-organization of fuzzy rule are realized by the self-learning of the neural networks. This approach is proven by experiments.

After introducing the techniques of application server and component, a system of electronic commerce based on Web is designed in use by combining the opening, cross-console and cross-language capabilities with the iPlanet middle-ware integrated framework. The structure of electronic commerce, application server and their security has been analyzed.