Dissertation > Industrial Technology > Automation technology,computer technology > Remote sensing technology > Interpretation, identification and processing of remote sensing images > Image processing methods

Research of Multi-source Image Matching Technology

Author FanManMan
Tutor JiaoBinLiang
School Yanshan University
Course Optical Engineering
Keywords remote sensing image registration SIFT algorithm Harris cornerdetection outliers removed
Type Master's thesis
Year 2012
Downloads 84
Quotes 0
Download Dissertation

The main purpose of image registration is to realize the process of geometricallyoverlaying two or more images of the same scene which are taken at different times,from different viewpoints or by different sensors. Remote sensing image registration isa crucial step in many image analysis tasks. It is popular used in environmentsurveillance, information fusion, machine vision, and other areas of research. In thispaper, many current registration techniques are referred. The multi-source remotesensing image registration techniques are further researched at the base of thesetechniques.The paper introduces scientific significance and application prospects of thistechnology detailly, also tells us the international research developments and theirtrends of this technology. A new algorithm based on multi-source-based point featurehigh-precision automatic satellite image registration which using from course to fineregistration strategy is proposed.First, the SIFT operator and a multinomial are used to achieve coarse imageregistration. The image after registrating at this time and reference images will be inthe same reference coordinate and the same scale,they are also in the same pixelsampling interval.Second, the Harris corner detection method is used to deal with the image to findthe point of feature. According the coordinates of the image, a smaller search range isdetermined. Then, referring to the correlation coefficients, the point with same namecan be got. However, the large data of remote sensing images and other factors such assurface features cover, shadows will affect the effect of this algorithm. The image willproduce inconsistencies locally and cause errors. So, it is needed to remove the errors.Compared with the traditional SIFT algorithm, the algorithm optimizes the featurepoint extracting time of traditional SIFT algorithm and avoids the low matchingefficiency problem when the image characteristics is uniformly distributed. Thematching experiment of SPOT satellite images are implemented with the new algorithm, and the results show that the new algorithm can enhance the matchingspeed effectively as the matching precision ensured.

Related Dissertations
More Dissertations