Application Of SIFT Matching Algorithm In Accuracy Evaluation Of Remote Sensing Image

Geographical conditions census is the prerequisite and basis of monitoring the geographical conditions. It has great significance for accurate information on the comprehensive national strength, improvement of the accuracy, regulatory, scientific and prospective of government management. It lays a solid foundation for promoting the construction of ecological civilization, social development, and the improvement of military power. It makes a great contribution to reaching a comprehensive well-off society. Remote sensing data which has a pivotal role in the first census geographic conditions is a major source of geographical conditions census. The accuracy of remote sensing image plane is the only indicator of the use of images. Since the sensor, time and spatial location is different, there are different image sensors, different time, different resolutions, different spectral phenomena, so it can not be used directly. In order to improve the efficiency and quality of the census, these images must be processed. The image indicators will meet the conditions of the existing census data and the latest data, information integration, removal redundant information, integration of complementary information, and ultimately be more accurate, richer, clearer image information than a single source of information.Because of the imaging mechanism of different sensors are different. The time, the angle, the environments of acquiring images are different. Though all feature information is in the same area, there is a difference in the electromagnetic spectrum, phase, and the spatial resolution. Therefore we need to deal with the image. Remote sensing image processing generally consists of image registration, image fusion, image mosaic, image segmentation and so on. No matter what kind of treatment method, the final quality inspection must be conducted by the inspection department. Therefore, remote sensing image qualified quality inspection is a necessary condition for the use of images. According to the national census DOM production technical regulations, test content includes (image plane precision, elevation accuracy, image color mode, image resolution, image quality, image cropping range, etc.) Whatever content, test methods are manual visual inspection, slow and inefficient. To improve the situation, the thesis is to study the imaging plane precision inspection method for example, in the end get the method of automatic inspection image plane precision.This paper is to study the remote sensing image data plane precision inspection method on geographic conditions census. For geographical conditions census accuracy of remote sensing image plane as an index, combining SIFT registration method, elaborates the algorithm principle and application of the theory. It mainly includes the principle, steps of image registration, mathematical model of image transformation (affine transformation) and the method of image registration. It starts from scale space to clarify the extraction methods of point feature in scale space, the importance and necessity of a Gaussian convolution in image registration. It has a detailed explanation of the principle of algorithm about the precise positioning of points in feature points extraction process, characteristics of rotation invariance, false matching points out and points out the advantages and disadvantages of this method of registration in practice which found that the number of feature points excessive and uneven distribution lead to matching accuracy is low and can not achieve the requirements of the census conditions. So from the distribution of feature points and the number of feature points to improve SIFT matching algorithm. Applying the improved SIFT registration method to the conditions census not only greatly improved the image matching speed, accuracy and saves manpower, material, financial resources for the census but also provided an efficient straightforward, high accuracy and other advantages, what’s more, improved the image of the plane edge accuracy assessment more convenience and higher accuracy.