The invention discloses a SAR image target recognition method based on a deep belief network. The method comprises the steps: (1) a to-be-recognized image sample is acquired; (2) filtering is performed; (3) characteristics are extracted; (4) the deep belief network DBN is trained; and (5) a target is recognized; the method adopts the deep belief network, various kinds of parameters are combined as the input, the problem of slow speed caused by amount of manual working when the single characteristic of the SAR image is adopted in prior art and cannot represent more comprehensive target information and utilize a surface learning method can be effectively solved, and the method can acquire more rich and more deep seated target information by combining various characteristics of the image and can improve SARS image target recognition precision.

SAR image target recognition method based on deep belief network

The invention discloses a living body face recognition method and device and equipment, and belongs to the field of face recognition. The method comprises the steps that an original image is obtained; according to the original image, a face image in the original image is obtained; PDAM illumination processing is carried out on the face image, and a first image is obtained; Fourier transform processing is carried out on the first image, and transform values of all pixel points of the first image are obtained; according to the transform values of all the pixel points of the first image and preset classification threshold values, whether the face image is the image of a living body face or not is determined. According to the living body face recognition method and device, living body face recognition accuracy can be improved with no auxiliary equipment or auxiliary operation of users, the calculated amount is small, operation is simple, and real-time performance is high.

Living body face recognition method and device and equipment

Islamic Azad University, South Tehran Branch, Electrical and Electronic EngineeringDepartment, Tehran, IranAbstract. We propose two methods for speckle suppression of synthetic aperture radar (SAR)images. The first method is based on Bayesian shrinkage and is a thresholding technique. ThemainproblemofapplyingBayesianshrinkageinatransformeddomain,suchascontourlettrans-form (CT), is finding the optimum threshold value. According to our experimental results, con-tourlet coefficients are affected by noise differently. It means that some contourlet coefficientsbelong to the specific sub-bands that are more robust against noise. We use this newfound prop-erty to determine the optimum threshold value and to develop our proposed method, which isnamed the weighted Bayesian shrinkage in contourlet domain. The second method, named theNSCT-GΓD, is a model-based approach using a two-sided generalized Gamma distribution(GΓD) to model the statistics of nonsubsampled contourlet transform (NSCT) coefficients.We use the Bayesian maximum a posteriori estimator to find NSCT despeckled coefficients.Experimental results carried out on both artificially speckled images and the true SAR imagesshow that our two proposed methods outperform other approaches via two point of views,speckle noise reduction and image quality preservation.

Two new methods based on contourlet transform for despeckling synthetic aperture radar images

A method for real face image recognition may include obtaining, by at least one processor, an human face image from an original image; obtaining, by at least one processor, a first image by executing a Point Divid Arithmetic Mean Illumination Treatment on the human face image; executing, by at least one processor, a Fourier transformation on the first image and obtaining, by at least one processor, the transformed value of each pixel of the first image; determining, by at least one processor, whether the human face image is a real human face according to the transformed value of each pixel of the first image and the preset classification threshold.

Systems and methods for real human face recognition

The invention provides an SVM-based background interference suppression method after radar target imaging. The SVM-based background interference suppression method comprises the steps of: firstly, performing signal preprocessing on a radar echo signal by adopting a moving target display (MTI) and mean removing method; secondly, performing two-dimensional Fourier transform on the pre-processed echosignal, and acquiring an RD (distance Doppler) diagram after radar target imaging; thirdly, regarding the generated RD diagram as a training sample and a test sample for training an SVM classification model; and finally, utilizing the generated SVM classification model to separate a plurality of targets and background interferences in a sample image, so as to realize background interference suppression. The SVM-based background interference suppression method is suitable for a variety of radar imaging scenarios, can effectively separate effective targets and background interferences, can effectively aggregate the targets while reserving the targets, can suppress the background interferences, and has broad application prospect.

SVM-based background interference suppression method after radar target imaging

The invention discloses a method for reducing speckle noises of an SAR image based on neighborhood directivity information, which mainly overcomes the defects that the prior method for reducing the speckle noises of the SAR image cannot inhibit the speckle noises well and lose part of marginal and detailed information. The method comprises the following steps: firstly, performing non-subsampled Contourlet decomposition on an original image, and estimating a binary mask to each high-frequency sub-band coefficient; secondly, obtaining an expression of a conditional likelihood ratio by using modified logarithmic Gaussian distribution and mixed exponential distribution according to corresponding different masks; thirdly, solving a priori ratio according to direction information of non-subsampled Contourlet transformation obtained by a neighborhood directivity model; and finally, obtaining a reduction factor according to the likelihood ratio and the priori ratio to reduce each high-frequency sub-band coefficient, and reconstructing the coefficient after the reduced change to obtain an output image after reducing the speckle noises. The method has the advantages of effectively eliminating the speckle noises of the SAR image and having good marginal retentivity, and can be applied to reducing the speckle noises of SAR images with abundant marginal information and abundant details, particularly the SAR images containing airports, runways and roads.

Method for reducing speckle noises of SAR image based on neighborhood directivity information

The invention discloses a method for recognizing an SAR target based on curvelet transform, and relates to the technical field of image processing and target recognition. The invention mainly solves the problems of low recognition rate and large volume of data existing in the prior method for recognizing the SAR target. The implementation process comprises the followings steps of: smoothing median filtering; performing nonlinear normalization; performing threshold segmentation after window filtering; extracting low-frequency subband information and the threshold segmentation by the curvelet transform; forming a fine characteristic area by morphological treatment; orderly performing the smooth filtering, the nonlinear normalization, the window filtering and the extraction of the low-frequency subband information of an image by the curvelet transform to form a rough characteristic area; combining the fine characteristic and the rough characteristic as a final target characteristic vector to train and support a vector classifier; and recognizing the SAR target by the well trained classifier. The method has the advantages of reducing the volume of data and improving the recognition rate, and can be applied to the identification of SAR ground targets.

Method for recognizing SAR target based on curvelet transform

An improved Nonsubsampled Contourlet Transform (NSCT)-based method has been proposed in this paper, using subbands mask prior models and directional information for synthetic aperture radar (SAR) image despeckling. The NSCT is a shift-invariant directional wavelet transform which satisfies the demands of SAR image despeckling. Wedding coefficient statistic character of NSCT subbands and directional information captured by the transform, this paper applies anisotropic neighborhood models to the processing of adaptive shrinkage. The resulting algorithm strongly suppresses speckle, while preserving detailed information and image edges compared with other traditional despeckling methods.

Hong Zhao

Papers

Method for reducing speckle noises of SAR image based on neighborhood directivity information

Method for recognizing SAR target based on curvelet transform

Adaptive shrinkage despeckling for SAR image based on Nonsubsampled Contourlet transform