Mathematical Morphology Combined With Invariant Moments For Remote Sensing Image Target Detection Method

With the rapid development of remote sensing technology, object detection technology of remote sensing images which was a project with value of both theory and practice was making great progress as an embranchment of remote sensing technology. Object detection technology of remote sensing images had been used widely in many fields of national defence and economic building. But traditional detection methods of linear and statistical theory didn't have met the new needs of object detection yet. Mathematical Morphology took great advantage over other spatial or frequency domain image processing and detection methods as a non-linear method based on geometrical features information. So this thesis designed and realized a detected arithmetic with matching method of object morphological template through the research on some key technology such as morphological feature extraction, object modeling and detection arithmetic. The central content and innovations in this thesis were followed as below.1.Expatiating the background and meaning of this thesis, analyzing the research actuality inside and outside country, summarizing the basic operations and natures of binary mathematical morphology and gray mathematical morphology.2.The foundational principle of edge detection was understood deeply and gripped at the base of summarizing and realizing classical edge detection operations agone. An improved self-adaptive multi-azimuth morphological structure element edge detection method was proposed and carried out by using multi-azimuth edge detection arithmetic through analyzing the development of morphological edge detection arithmetic.3.Hu moment, Zernike moment and Hausdorff distance were studied assembled. The invariable features of Hu moment and Zernike moment was analyzed and experiments corresponding were done. Applying Hu moment and Zernike moment to Hausdorff distance to select template for detecting image object. At the precondition of estimating the characteristics of target image, threshold could be set to control the precision so as to reduce error detections and avoid needless inaccuracy.4.The whole flow of simple object detection was realized by combining the Mathematical Morphology and invariable feature arithmetic rationally. This method was of great merit such as small amount of computation, easy realization through hardware, high detection precision, capacity of anti-noise, better practicability and wide application etc.