Search

On the basis of grey theory model for medium-long term load forecasting, the paper improves the traditional grey model and brings forth the grey residual forecasting model for less prediction errors. At the same time, it uses fixed weight method of forecasting model--that is to choose weight with average precision of each model to build a optimization combination forecasting model. It forecasts with the historical load data of some area, and proofs that combination forecasting method is superior to single forecasting model, Its forecasting error is less than 1%. Finally it forecasts future five year loads of some area, and provides a tech support and theoretical foundation for area Power Company.

The intention of this paper is to help bridging the gap between knowledge base and computer vision system. A knowledge-based vision system for identification of overlapping objects is presented. The authors place emphasis on the reasoning strategy based on knowledge base for recognizing of occluded workpieces to provide information with an education Robot. The experimental results are given and some problem are discussed.

This paper presents an improved stereo matching algorithm by heuristic strategy technique resulting from that the stereo matching problem can be cast as a search one. After early processing, our algorithm employs two search ways. One is intra-scanline search and the other a inter-scanline search. Both of them a can be calculated separately and treated as the problem of finding an optimal path on 2-D search graph by graph searching method with the concentration on minimizing some combination of cost of the path and effor of the search required to get the path, that is heuristic. Our algorithm hs been tested with block scences and its computational requirements have been discussed too.

A novel method for spectral similarity measure, which is called nonlinear spectral similarity measure, is presented in This work. In this method, all original spectral vectors are, firstly, nonlinearly transformed into a feature space. Next, kernel PCA is used to construct a set of orthogonal coordinate base in feature space. All transformed spectral vectors are projected onto the orthogonal coordinate space. In kernel principal component analysis (KPCA), the nonlinear translation function is implicatively implemented by kernel function. Moreover, all projected spectral vectors are constrained by spectral continuum removal curve. Because of continuum removal curve, various bands contribute the similar measurement differently. The more absorption, the more the contribution in similarity measurement. At last, linear or general linear similarity measure, for example spectral angle mapper, was used to measure the similarity between two nonlinearly transformed spectra. Our experiments show that this method is effective in spectral similarity measure.

A new method for natural color image segmentation using integrated features is proposed in this paper. Edges are first detected in term of the high phase congruency in the gray-level image. K-means cluster is used to label long edge lines based on the global color information to estimate roughly the distribution of objects in the image, while short ones are merged based on their positions and local color differences to eliminate the negative affection caused by texture or other trivial features in image. Region growing technique is employed to achieve the final segmentation results. The proposed method unifies edges, both the whole and local color distributions, as well as the spatial information to solve the natural image segmentation problem. The feasibility and effectiveness of this method have been demonstrated by various experiments.

Vehicle detection is one of the key technologies in intelligent transportation system (ITS), and it is an important stage of vehicle tracking in visual surveillance. Due to the clutter of traffic scenes, the captured video sequence involves many large noises. In analyzing human vision, the non-object area and shape suppression are defined, and a new noise-removing algorithm is proposed. To reduce the computational complexity, the sequential algorithm is improved. Experimental results show the effectiveness and efficiency of the algorithm in extracting the moving vehicles in a cluttered scene. The output of the vehicle's contour is very complete and accurate, and can be used in vehicle tracking system to improve the performance.

Digital watermarks have been proposed for the copyright protection of multimedia products. Most schemes were performed primarily for fidelity considerations, and not specifically to improve the detector performance. In this paper, a robust and blind watermarking scheme is presented. The concept of communication with side information was applied at the encoder to improve the probability of detection within acceptable fidelity, and the template matching technique was employed to estimate the undergone attacks in attacking channel. This scheme is optimized by optimal design of the encoder to match with the media content and the decoder to adapt to the attack channel state. Experiments show that our method is robust against some common attacks such as filtering, compression, rotation, scale, and translation. It can be applied to both color and gray images.

Video object (VO) is an important concept in MPEG-4. For objects can be easily manipulated without visible distortion, the copyright protection of video objects becomes an important issue. This paper presents a watermarking scheme for video objects. Different from other methods, the proposed scheme employed inertia ellipse to achieve fast synchronization recovery in case the object was manipulated. Shape adaptive DCT and visual mask were combined to embed the watermark into the arbitrarily shaped object, which was designed to achieve the trade-off between the invisibility and the robustness. Experiments show that our algorithm is robust to object manipulations such as rotations, translations, scaling and lossy compression. Our scheme can be easily incorporated into the object-based coding framework of MPEG-4.

We present a vision system that can be to recognize 3D objects. Surfaces are characterized by surface curvature. Qualitative features are shown to be effective in pruning the search space of possible scene interpretations explored by the recognition and localization technique. Within this paradigm, we implemented a system which extracts planar patches, and patches of quadrics of revolution from a depth map. Reconstructed features are indexed into a model database to verify object hypotheses using the Hough technique. Our approach is surface based and is not sensitive to noise or occlusion. >