Multi-features Remote Sensing Forest Vegetation Segmentation Based On The Theory Of Blue Noise

Recognition of forest type is an important work for remote sensing detection of forest resources.High resolution remote sensing image not only contains abundant spectral information, also provides a rich space and texture feature information,so, utilizing high resolution remote sensing image to identify the features and forest tree species is become the focus of research in recent years.In allusion of the high resolution images have the great amount of data and the features of various types,using single feature segmentation has great limitations,It cannot fully take advantage of image information, such as low classification accuracy, slow speed limitations, also can cause a lot of spatial data redundancy and waste of resources.Therefore, the method combining multi-features of image segmentation has become a focus of research.To solve this problem mentioned above,a combination method of multi-scale gray, shapes and textural features has come up.At first,this method uses the normalized difference vegetation index to extract vegetation of image,according to the vegetation spectral radiation characteristics.Then,the way extracts multi-scales of forest vegetation,basing on the blue noise theory.Finally,the way to obtain final segmentation consequence is to build crown constructing primitive at different scale with combination of gray level, shape and other texture features,then utilizing that constructing primitive to extract original image.In the process of extraction,the design of the structure element of forest texture is particularly important.In this paper, on the basis of morphological structure elements,by analyzing micro structural typical vegetation texture,based on multi-scale information,we build structure element to describe the forest vegetation.On the basis of theory research above,remote sensing image is binding data of Quick Bird,the panchromatic high-resolution is 0.65 m,multi-spectral resolution is 2.76 m,in view of the blue noise feature extraction.The experimental result indicates that arithmetic in the article could advance the veracity of vegetation field with a preferable segmentation effect.