Remote Sensing Image Segmentation Algorithm Based On Multiple Features

With the improvement of remote sensing satellite resolution,remote sensing technology has been widely used in the fields of land and resources utilization,urban environmental monitoring and natural disaster prediction.Segmentation is not only an important aspect of remote sensing image application,but also one of the key points and hotspots of remote sensing image research.Generally speaking,remote sensing images contain a large number of features,like spectral features,shape features and texture features.If only one of these features is used,it is usually difficult to obtain high-precision segmentation results.Therefore,how to reasonably use these different features to segment remote sensing images and obtain better segmentation results is worth studying.This paper studies the segmentation of remote sensing image from the aspects of feature extraction,feature selection and feature fusion.The main work of this paper is given as follows:1.A remote sensing image segmentation algorithm based on ant colony optimization and multi-class feature fusion is proposed.Firstly,the path selection mechanism and pheromone update mechanism in binary ant colony algorithm are improved,and a feature selection method based on improved binary ant colony algorithm is proposed.The horizontal fusion of multiple types of features is completed by feature selection of multiple types of features,and finally the best fusion features obtained are used for image segmentation.The experimental results on synthetic texture images,synthetic remote sensing images and real remote sensing images show that: 1)for most images,the effect of multi-class feature fusion is better than that of single feature;2)the feature selection method based on improved binary ant colony algorithm is more effective than the other five feature selection algorithms;3)the proposed algorithm can adaptively select the most suitable fusion features for different images and achieve better segmentation results.2.A remote sensing image segmentation algorithm based on multi-class feature fusion based on adaptive weight is proposed.In order to solve the problem that multi-class feature fusion does not consider the influence of feature weight on segmentation results in the first work,the algorithm uses feature selection method to complete multi-class feature fusion,then genetic algorithm is used to optimize all kinds of feature weights of fusion features,and penalty method and repair method are used to deal with constraints respectively.The experimental results on synthetic texture images,synthetic remote sensing images and real remote sensing images show that: the proposed algorithm can adaptively select the best fusion features for different images and assign the most appropriate feature weights.Compared with the fusion features without weight optimization,the segmentation result of the proposed algorithm is improved to a certain extent.3.A remote sensing image segmentation algorithm based on ant colony optimization multidimensional feature selection is proposed.The essence of the algorithm is to select each dimensional feature on the original feature set and add a piecewise mutation strategy to the improved binary ant colony algorithm.The experimental results on synthetic texture images,synthetic remote sensing images and real remote sensing images show that: 1)the proposed algorithm has a good effect on improving the segmentation performance and reducing the feature dimension;2)compared with the other two kinds of multi-features,the segmentation result of using multi-dimensional feature selection is better.Because this method can select useful features in finer dimensions,delete irrelevant or redundant features,and make full use of the rich information of multiple features.