Research And Application Of Hyperspectral Remote Sensing Image Target Detection Method

Due to the low spatial resolution of hyperspectral remote sensing images,the target objects in the image are often weak information states,so that the single target detection method based on spatial morphology can not solve the problem of target detection of hyperspectral remote sensing images.In this paper,we use the hyperspectral remote sensing image to enrich the spectral information,analyze the difference of spectral features in the target features to detect and identify,and propose two new methods of target detection.The hyperspectral remote sensing image target detection based on the sparse expression model is obtained by using the classical adaptive cosine consistency estimator algorithm to reduce the image dimension and to combine the spectral information of the target object with the image.The second method is to introduce the kernel method,apply the deformed Gaussian kernel function to the hyperspectral remote sensing image target detection,use the kernel function to map the original input data to the high dimensional feature space,and apply the constraint of the weighted sample covariance matrix in the high dimension feature space Energy minimization linear algorithm,the separation of the target and the background has achieved good results.In this paper,the experimental results of different sparse levels L in the hyperspectral remote sensing image target detection method based on the sparse expression model are analyzed,and the optimal sparse degree of the target detection effect is obtained.Both methods are compared with the results of a single traditional target detection algorithm,and the target detection rate is significantly improved.And the two kinds of target detection with the same group of data for the experiment,the results of comparative analysis of the advantages and disadvantages of the algorithm.