Wavelet transform, multi-strategy PSO (particle swarm optimization) and SVM (support vector machine) integrated based remote sensing image classification method

Abstract

The invention relates to a wavelet transform, multi-strategy PSO (particle swarm optimization) and SVM (support vector machine) integrated based remote sensing image classification method. The method includes the following steps of 1), optionally selecting a remote sensing image to be classified, subjecting the image to grey processing and transforming the same into a corresponding grey image; 2), subjecting the grey image to noise suppressing preprocessing to acquire a preprocessed remote sensing image; 3), subjecting the preprocessed remote image to textural feature extraction by adopting wavelet transform prior to normalization processing to acquire textural feature vectors of the remote sensing image; 4), realizing the wavelet transform, multi-strategy PSO and SVM integrated based remote sensing image classification method by adopting a multi-strategy improved particle swarm optimization algorithm and parameters used for optimizing a SVM classifier, and classifying the textural feature vectors of the remote image to be classified to acquire attributes of the remote image. Therefore, the wavelet transform, multi-strategy PSO and SVM integrated based remote sensing image classification method is widely applicable to the technical field of computer image retrieval.

Description

technical field [0001] The present invention relates to a remote sensing image classification method, in particular to a remote sensing image classification method based on the integration of wavelet transform, multi-strategy PSO (Particle Swarm Optimization, particle swarm optimization algorithm) and SVM (Support Vector Machine, support vector machine). Background technique [0002] Remote sensing image classification is an important tool for remote sensing data analysis and information extraction. It is essentially a pattern recognition technology including feature extraction and classifier design. The purpose is to convert remote sensing images into usable geographic data. It is regarded as a remote sensing research field. An extremely important research content. Due to the large number, variety, large gray scale, fuzzy boundary, illumination, angle, magnitude, and small samples of remote sensing images, remote sensing image classification has always been a hot and diffic...

AI Technical Summary

Problems solved by technology

These image classification methods based on SVM and its improvements have achieved better classification results, but there are still image representations and feature vectors that are interfered by other pixels, resulting in incomplete feature extraction, and the increase in the number of training samples leads to reduced classification efficiency and longer time, etc. problem; and the classification accuracy is related to the SVM classifier. For a given error parameter, the subjectivity is too strong, which is easy to cause subjective errors, making it difficult for the SVM classifier to obtain better classification performance.

At the same time, with the significant increase in the amount of remote sensing data and the highly detailed information, the data dimension becomes larger and larger. Under the condition that the image classification results and accuracy requirements are the same, the image processing time and cost increase significantly. It has become quite difficult. When using the current SVM classifier for image classification or recognition, it is impossible to make a choice on the effectiveness of image features, and it is difficult to extract complete remote sensing image features, which makes the classification results difficult to meet the actual remote sensing image classification requirements.

Images