Research Of Content-Based Satellite Cloud Image Processing And Information Retrieval Technology

Recently, with the coming of information age especially digital age, the computer information processing has become most important tool for satellite cloud image processing. Satellite cloud image is one of the remote sense data sources which include abundant information. The modern weather forecast mainstream work pattern is like this: using image processing method analyzed and extracted useful image character and content, then judge the atmosphere change and weather situation. So it has become the main research direction of satellite cloud image processing for how to process, analyze and apply satellite cloud image information.In this thesis, using traditional image processing method for reference, bring forward content-based image processing and information retrieval application method, raising purpose and efficiency of image processing, it has better innovation and application value. The image content put forward here includes two layers of meaning: firstly, it is the basic image character like texture, color, shape and edge, etc; secondly, it means the information that character has reflected, that is the weather condition or development trend describer by the cloud image series. In this thesis, analyzed the two layers of image content systematically, aiming at image content, using several image character extraction algorithm, then analyzed the weather information reflected by image character, reach the research purpose of practical application.Concretely, in this thesis, the research achievement and main work includes:1. Introduce the content-based processing theory in satellite cloud image processing fieldIn this thesis, as the main research content, the character of satellite cloud image was studied from bottom layer character extraction and high level image retrieval two directions. And then the image undersatand category was introduced in satellite cloud image processing field. 2. Raised the recognition criterion of satellite cloud image comprehension base on the satellite cloud image characterOn the basis of abundant satellite cloud image analysis, raised general cloud image recognition criterion, and acquired four basic image character (color, texture, edge and shape) from the six criteria, establish the foundation for cloud processing and recognition accurately.3. Raised the satellite cloud image filtering algorithm based on integration filtering theory in wavelet fieldSince the imaging theory, the resolution and definition of remote sensing image still need to be raised, the image "soft" processing technology has become very important in the condition of low hardware level. The integration filtering method base on wavelet field use the smooth filtering theory in wavelet decompose image, it could reduce disturb in wavelet field, the quality of reconstruction image has achieved obvious improvement.4. Analyzed the character of satellite cloud image systemtically, raised improvable edge detection fusion algorithmThe basic charcter of satellite cloud image could be devided into spectral character, texture character and spatial geometry character. Begin from color processing and gray correction, the color character was studied, improve the image quality; the texture was analyzed systematically, using co-occurrence matrix method with statistical feature analyze the image texture character parameter, realized the character recognition of satellite cloud image. Especially aim at the edge character analysis of satellite cloud image, combining with traditional sobel and canny edge detector, raised amendatory edge detection fusion algorithm. Base on strong edge detection theory and three optimal detection criteria, the algorithm use edge accurate positioning and weak edge modification theory, solve the adaptive threshold chosen question well, achieved better effect.5. Raised the region grow algorithm base on character vector space and competition grow theory The region character extraction of satellite cloud image is always a difficult problem; we often face some questions like "missing" or"flow". In this thesis, improve the general region grow algorithm, describe the image in vector space, raised three grow criteria which include whole compare detection, acerage measyre and space first, equal competion oppptunity, etc. using label marking method decide the position of character vector, and we could acquire the object accurately.6. Base on varational method, raised two kinds of satellite cloud image contour character extraction methodThe contour character is a very important visual character, in this thesis, using varational theory acquire satellite cloud contour. The two big branch ofvarational theory was studied------parameter active contour model and gemmetricactive contour model, two kind of contour extraction algorithm was raised. The algorithm colud acquire satellite cloud contour accurately, which has strong reliability and practicability.7. Raised a efficient image multi-character weighted value feedback retrieval algorithm adapt to satellite cloud imageIn this thesis, the whole process of satellite cloud image retreival was studied, which includes four steps like character extraction, character index, image matching and character feedback. The research begins from image character like color, texture, shape and edge, the image retrieval method base on different character was studied synthetically, raised an efficient image weighted value feedback retrieval algorithm, which could retrieve correlative image information accurately and efficiently, reach the research aim of make the best of the satellite cloud image database, provided a better application platform for user.On the basis of analyzed quantities of image and use traditional method for reference, aim at satellite cloud image character, using several adaptive satellite cloud image processing method, and then achieved better experiment result. Since the weather information quantity is so big, the adaptable satellite cloud image database was constructed and content-based image retrieval method was studied for quick search and accurate analysis. The research combine the whole operation process of satellite cloud image processing, which includes processing, analysis, storage, retrieval, application and forecast, enhanced the application efficiency of cloud image, and it has better theory value and application value.