The Research Of An Image Retrieval Technology For Stock Analysis

Trend Analysis has important meaning in the reality, either to national economic situation analysis greatly, or to the weather forecast of month, or in the sale quantity of cold drink store the next day. It can make a very import role in the stock certificate, because of the large amount of public information.The article mainly inquiry into how to makes use of calculator image processing to analysis the relation between the stock price trend in future and the trend ago and technique index sign gao to give the investor some certain reference.The main research work of the article is to discuss the possibility to makes use of some image processing's technique to carry on trend analysis and for this purpose what kind of image processing technigue can be revolvedChapter 1,Introduction, to introduced some research background among them;Chapter 2, Trend analysis and image processing technique can be used;Chapter 3,Characteristic inquiring based on shape withdraws;Chapter 4, Characteristic inquiring based on object space withdraws which means relation between objects, chapter 5 Comparison method which inquiry into image likeness-degree;Chapter 6, Multiple regression analysis, try to study the relation between the technique index sign of the assurance stock certificates.Chapter 7,Summary and outlook.With the rapid development of the multimedia network technology, the application of the image becomes more and more extensive and Content-based Image Retrieval (CBIR) becomes one of most active researches in multimedia retrieval field. In order to analyze the information of an image, the CBIR system always analyzes the color, texture, shape, and other low-level image features, taking feature vectors as retrieval index. Up to now, the main CBIR method is similarity matching based on multi-dimension feature vector of image. Extracting features from image and similarity match are the key issues in CBIR.In this paper, we extensively studied the national and international materials on CBIR systems, discuss the research status and trend in content based image retrieval problems, and analyzed and implemented some methods of extracting image content in details. In this paper, how to extract the feature of shape and space-interrelation and how to integrate the feature of color, texture, shape and space will be discussed detail. This dissertation discusses the visual content based method based on image feature extraction and similarity matching. We designs an image retrieval system based on image content, the experiments show that the shape features and thespace-interrelation features are effective in describing image content, and the combined feature by the four feature is superior to any single feature of the color, texture, shape and space-interrelation on retrieval.Shape is another important feature of object in image, has good stability. In the retrieval mainly based on similar shape feature, the shape feature shows the fine capability, which the color and the texture feature can't compare with. So the research on how to extract shape feature from image is very important. This paper has analyzed the shape feature and presented a moment invariant based on threshold optimization method. In this method, we transformed the color image to gray image first, and used the moment invariant based on threshold optimization to segment the gray image, and then extracted the edge of gray image. So far as we have obtained the edge of gray image, and calculates the seven-moment invariant of edge of gray image. The seven-moment invariant was the shape feature of the image. In order to make the seven-moment invariant supply the same measure, we should be make every the seven-moment invariant dimensional uniformity through Gauss-uniformity. The result of experiment shows that moment-invariant possess the image object's invariant to translation, scaling and rotation, it can describe the shape of image and the space information well, and especially for the image with clear edges and object the retrieval result is good.Spatial relationship is the important feature of object in image and it describes the inter-relationship between every object in image. When the shape and size of object is far less than the distance of the two objects, the retrieval by spatial relationship is more effective. In this paper, a new content-based spatial relationship retrieval method is proposed. In this method, we translate the RGB space of color image into HSV space first, and then quantifies color sector with unequal interval, and get feature vector, take every bin of H as a state, then we called the gray image of H as state matrix. Using the Z-scans, the state matrix is transformed into 1-dimensional state sequence. Here we suppose that the 1-dimensional state sequence accords with the Markov chains. Then the 1-step transition probability matrix of the sequence is calculated as the image's spatial distributions information. The 1-step transition probability matrix is the feature of spatial relationship of image, and then we can use it in retrieval.The color, texture, shape and spatial relationship only describe one part of image content. Single feature cannot present all the contents of image. In this dissertation a method used feature of combined color, texture, shape and spatial relationship is also given. In order to combine the four feature together, we should make the four feature normalized, so that they have the same dimension. According to the capability ofretrieval, we distribute the weight to every feature, and then combine four-feature together. Experiment shows that combined feature is superior to every single one of the four features in retrieval.In this dissertation, we implement an automatic system of extracting color, texture, shape features from image based on content and an image retrieval demo system as the platform for extracting feature algorithms, which is an experimental framework system. The system for developing this system is Windows, and the development environment is Visual C++ 6.0 and Microsoft SQL Server. Further work will be done to develop this system so as to release an applied commercial version. This dissertation holds certain referential value and practical significance in promoting the development of retrieval technique of image database.