Knowledge-Based Restriction Segmentation Method For High Spatial Resolution Remote Sensing Images

With the development of remote sensing technology and the acquisition of high spatial resolution remote sensing images,object-based image analysis has become one of the most popular remote sensing image information extraction methods,and one of the key technologies of object-based image analysis is remote sensing image segmentation.Remote sensing image segmentation is the basis for image recognition and analysis.The accuracy of remote sensing image segmentation directly affects the accuracy of subsequent image understanding and application.Therefore,remote sensing image segmentation plays a important role in object-based image analysis.However,in the current remote sensing image segmentation,there are still several problems: firstly,the segmentation effect is easily affected by noise and other factors;secondly,the problem of segmentation of object boundary accuracy;thirdly,the problem of optimal segmentation scale,if the scale parameter is too large,there is maybe cause undersegmentation,and if the scale parameter is too small,the over-segmentation problem is easily caused.The root cause of these problems is the difficulty of integrating knowledge effectively in the segmentation process.In response to these problems,and taking into account the particularity of remote sensing image segmentation,this paper classifies the knowledge in remote sensing image segmentation and divides it into internal knowledge and external knowledge,and proposes high spatial resolution remote sensing image fine segmentation method based on internal knowledge and external knowledge constrained,respectively.For the edge constraint problem in internal knowledge,this paper proposes a watershed segmentation algorithm and a region growing segmentation algorithm for high spatial resolution remote sensing images based on edge constraints.Experimental verification this paper proposes that the segmentation method can guarantee the integrity of the segmentation patch to the maximum extent and can suppress the undersegmentation of the image.Aiming at the scale constraints in the internal knowledge,this paper designs a high spatial resolution remote sensing image segmentation algorithm based on the optimal scale constraints.Based on the method of local variance statistics,this paper proposes a segmentation scale estimation method based on watershed segmentation algorithm and analyzes the reliability of the method.For external knowledge constraints,external knowledge transfer applied to image segmentation process.This paper proposes a high spatial resolution remote sensing image segmentation method based on constraints of land use maps.The reliability and rationality of the method are verified by experiments.This paper systematically elaborates the knowledge and its classification of remote sensing image segmentation,and applies it to image segmentation algorithms.It effectively controls and constrains the segmentation results,guarantees the segmentation accuracy to a certain degree,and provides good image data foundation for subsequent object-oriented analysis and application.It also provides theoretical and methodological guidance and reference for the study of image segmentation theory.