Hybrid sorting process of adjustable spectrum and space characteristics

Abstract

An adjustable mixed classification method of spectrum and space character includes setting a series of multimeasure windows of central pixel and using Mallat transform to set up multiresolution analysis of each window, obtaining four wavelet subbands by each time of Mallat decomposition on each window, using four said subbands to structure a rotary unchanged space character, selecting sample similar to said type and using mean value and variance measure to structure deflection vector according to spectrum and space information of sample, making character weight on different space character and selecting occupied proportion of spectrum and space character in classification.

Description

technical field [0001] The invention belongs to the technical field of computer remote sensing image processing and pattern recognition, and is a new classification method for extracting image space features by using multi-scale background and multi-resolution analysis, and adjusting the ratio of spectrum and space information in a classifier. Background technique [0002] In recent years, the development of space technology has enabled us to obtain remote sensing data with higher spatial resolution. The characteristics of this image are: (1) adjacent pixels have a high degree of correlation; (2) the spectral difference within the class increases, (3) The spatial relationship of the image becomes more complex, and the phenomena of the same spectrum with different properties and the same mass with different spectra widely exist. Therefore, there is a large overlap in the spectral information of different types of ground objects, and traditional spectral methods face difficult...

AI Technical Summary

Problems solved by technology

It must be pointed out that this kind of window selection is only a compromise solution for actual processing. Due to the multi-scale characteristics of image features, it is difficult to effectively express and describe them with one window scale.

[0010] 2. The existing wavelet method can effectively identify certain texture features, but its shortcomings are: (1) It does not solve the window problem described above. The algorithm uses a larger fixed window to operate, and often can only detect a larger range. The texture (such as large-scale landforms) cannot be applied to the actual remote sensing image recognition; (2) wavelet multi-resolution analysis will produce high-dimensional spatial features, these features have a lot of information redundancy, and some features have been proved to be useless Therefore, it is necessary to extract features from a large number of wavelet sub-band energy values; (3) the wavelet texture features and spectral information are not combined, especially in high-resolution remote sensing images, the combination of spectral and texture information It is crucial for the classification of images

Images