Research On Hyperspectral Image Dimension Reduction Method Based On Artificial Fish Swarm Algorithm

In the field of remote sensing(RS),hyperspectral imaging(HSI)is a key technology and a major breakthrough in the history of remote sensing technology.The main principle is that there are a lot of closely adjacent electromagnetic waves from remote sensing satellites.The reflection of these electromagnetic waves on objects on the earth will be received by the sensors on the satellite,and the computer will obtain and record the information.There will be too much information in the original data band,so there will be too many electromagnetic sensors.However,these hyperspectral image data contain not only useful data,but also a lot of redundant data.A large number of useless data will lead to the difficulties of researchers in post-processing images,and these unnecessary information will also lead to Hughes phenomenon.In view of the above problems,the existing solutions can be roughly divided into two categories,one is the means of feature selection,the other is the method of feature extraction.The ultimate goal of both methods is to reduce the amount of hyperspectral image data and extract useful information from it by reducing the dimension of the original data set.This paper mainly uses the method of band selection.As a common hyperspectral image processing technology,the best band hidden in the original data set can be separated by some technical means.This technical means can effectively remove redundant information,effectively reduce the workload of later technicians,and show excellent performance in the experimental results Classification results.In order to eliminate the dimensionality disaster and redundancy of hyperspectral data,based on the traditional Subspace Partition Technology and artificial fish swarm algorithm(AFSA),this paper analyzes the known dimensionality reduction algorithms and explores new solutions,and proposes a new band selection method based on fuzzy c-means clustering artificial fish swarm algorithm(fcm-fs),which acts on three criteria in hyperspectral image experiment Standardized data set.Compared with the traditional subspace partition method,this method can roughly determine the number of Subspace Partition with the support of hyperspectral image spectral visualization technology,and use FCM algorithm to effectively divide the similar bands in the original data set into the same subspace,and then through experiments and research,it is determined that the maximum entropy is the fitness function of the artificial fish swarm optimization algorithm One idea is to select the band from each subspace.In order to illustrate the effectiveness of the proposed algorithm,other commonly used band selection algorithms are found to act on the three hyperspectral standard data sets.The fcm-fs algorithm proposed in this paper is compared with other algorithms.The results show that the proposed algorithm can effectively reduce the redundancy between the selected features and maintain high classification performance.