Zhirong Men

Bio: Zhirong Men is an academic researcher from Beihang University. The author has contributed to research in topics: Synthetic aperture radar & Azimuth. The author has an hindex of 6, co-authored 24 publications receiving 126 citations.

SAR ATR based on dividing CNN into CAE and SNN

Abstract: As for the problem of too long training time of convolution neural network (CNN), this paper proposes a fast training method for CNN in SAR automatic target recognition (ATR). The CNN is divided into two parts: one that contains all the convolution layers and sub-sampling layers is considered as convolutional auto-encoder (CAE) for unsupervised training to extract high-level features; the other that contains fully connected layers is regarded as shallow neural network (SNN) to work as a classifier. The experiment based on MSATR database shows that the proposed method can tremendously reduce the training time with little loss of recognition rate.

Multi-view SAR image target recognition method based on depth neural network

Abstract: The invention discloses a multi-view SAR image target recognition method based on a depth neural network, and the method comprises three steps: image preprocessing, feature extraction based on CAE, and multi-view SAR image recognition based on RNN. The method specifically comprises the steps: firstly carrying out the cutting and energy normalization of an inputted image; secondly extracting the features of an original image through the non-supervision training of CAE; thirdly constructing a multi-view SAR image feature sequence through the above features; fourthly carrying out the supervised training of the RNN through employing a training feature sequence, wherein the RNN can be used for the recognition of a testing set feature sequence after training. The method can make the most of the capability of CNN in learning and extracting the general features of the image and the capability of RNN in fully extracting the context of the sequence, effectively improves the recognition rate of a multi-view SAR image target, and is higher in engineering value.

Geometric correction method for spotlight-mode satellite SAR (synthetic aperture radar) image

Abstract: The invention discloses a geometric correction method for a spotlight-mode satellite SAR image. The method comprises the following steps: (1) reading related parameters and echo data, and constructing a positioning equation; (2) calculating the coordinates of the position vector and velocity vector of a satellite in a non-rotating geocentric coordinate system at a moment of radar illumination center according to the spatial geometric relationship between the satellite and the earth; (3) completing slope distance and Doppler frequency correction; (4) acquiring the coordinates as well as longitude and latitude of a certain pixel point (i, j) of the satellite SAR image in the non-rotating coordinate system; (5) determining whether all of the pixel points of the satellite SAR image are processed; and (6) re-dividing the latitude-longitude grid, and outputting the geometric correction result. The method provided by the invention can be used for carrying out row-by-row geometric correction on the satellite SAR image, and can process the satellite SAR image in blocks as required due to the completely independent treatment process between the rows of the satellite SAR image, thereby further improving the treatment efficiency.

A Modified Radon Fourier Transform for GNSS-Based Bistatic Radar Target Detection

Abstract: The Global Navigation Satellite System (GNSS)-based passive bistatic radar (PBR) which uses the GNSS signal as the illuminators of opportunity is studied for moving target detection (MTD). GNSS-based PBR has many advantages due to the removal of the transmitting device; however, its fundamental limitation is the low power density of the GNSS signal. Therefore, the integration time should be sufficiently long to obtain a promising maximum detectable range. On the other hand, the integration time is limited by the range migration and Doppler migration of the echo caused by target motion. In this letter, a novel MTD algorithm is proposed for the GNSS-based PBR, by employing a modified radon Fourier transform (MRFT) to achieve the required long-time integration for moving targets. The MRFT integrates the echo energy via joint searching of range, Doppler, and Doppler rate of the target, which can handle not only the range migration but also the Doppler migration problems, and significantly improves the signal-to-noise ratio (SNR) of the echo signal. An experiment using the GPS L5 signal as the illumination source is conducted and a moving car is successfully detected by the proposed algorithm, although significant range migration and Doppler migration are present due to variation of its speed.

High-Temporal-Resolution High-Spatial-Resolution Spaceborne SAR Based on Continuously Varying PRF

Abstract: Synthetic Aperture Radar (SAR) is a well-established and powerful imaging technique for acquiring high-spatial-resolution images of the Earth's surface. With the development of beam steering techniques, sliding spotlight and staring spotlight modes have been employed to support high-spatial-resolution applications. In addition to this strengthened high-spatial-resolution and wide-swath capability, high-temporal-resolution (short repeat-observation interval) represents a key capability for numerous applications. However, conventional SAR systems are limited in that the same patch can only be illuminated for several seconds within a single pass. This paper considers a novel high-squint-angle system intended to acquire high-spatial-resolution spaceborne SAR images with repeat-observation intervals varying from tens of seconds to several minutes within a single pass. However, an exponentially increased range cell migration would arise and lead to a conflict between the receive window and 'blind ranges'. An efficient data acquisition technique for high-temporal-resolution, high-spatial-resolution and high-squint-angle spaceborne SAR, in which the pulse repetition frequency (PRF) is continuously varied according to the changing slant range, is presented in this paper. This technique allows echo data to remain in the receive window instead of conflicting with the transmitted pulse or nadir echo. Considering the precision of hardware, a compromise and practical strategy is also proposed. Furthermore, a detailed performance analysis of range ambiguities is provided with respect to parameters of TerraSAR-X. For strong point-like targets, the range ambiguity of this technique would be better than that of uniform PRF technique. For this innovative technique, a resampling strategy and modified imaging algorithm have been developed to handle the non-uniformly sampled echo data. Simulations are performed to validate the efficiency of the proposed technique and the associated imaging algorithm.

A New Algorithm for SAR Image Target Recognition Based on an Improved Deep Convolutional Neural Network

Abstract: In an attempt to exploit the automatic feature extraction ability of biologically-inspired deep learning models, and enhance the learning of target features, we propose a novel deep learning algorithm. This is based on a deep convolutional neural network (DCNN) trained with an improved cost function, and combined with a support vector machine (SVM). Specifically, class separation information, which explicitly facilitates intra-class compactness and inter-class separability in the process of learning features, is added to an improved cost function as a regularization term, to enhance the DCNN’s feature extraction ability. The enhanced DCNN is applied to learn the features of Synthetic Aperture Radar (SAR) images, and the SVM is utilized to map features into output labels. Simulation experiments are performed using benchmark SAR image data from the Moving and Stationary Target Acquisition and Recognition (MSTAR) database. Comparative results demonstrate the effectiveness of our proposed method, with an average accuracy of 99% on ten types of targets, including variants and articulated targets. We conclude that our proposed DCNN method has significant potential to be exploited for SAR image target recognition, and can serve as a new benchmark for the research community.

SAR ATR based on displacement- and rotation-insensitive CNN

Abstract: Among many synthetic aperture radar (SAR) automatic target recognition (ATR) algorithms, convolutional neural network (CNN)-based algorithms are the commonly used methods. However, most previous SA...

POLSAR Image Classification via Wishart-AE Model or Wishart-CAE Model

Abstract: Neural network such as an autoencoder (AE) and a convolutional autoencoder (CAE) have been successfully applied in image feature extraction. For the statistical distribution of polarimetric synthetic aperture radar (POLSAR) data, we combine the Wishart distance measurement into the training process of the AE and the CAE. In this paper, a new type of AE and CAE is specially defined, which we name them Wishart-AE (WAE) and Wishart-CAE (WCAE). Furthermore, we connect the WAE or the WCAE with a softmax classifier to compose a classification model for the purpose of POLSAR image classification. Compared with AE and CAE models, WAE and WCAE models can achieve higher classification accuracy because they could obtain the classification features, which are more suitable for POLSAR data. What is more, the WCAE model utilizes the local spatial information of a POLSAR image when compared with the WAE model. A convolutional natural network (CNN), which also makes use of the spatial information, has been widely applied in image classification, but our WCAE model is time-saving than the CNN model. Given the above, our methods not only improve the classification performance but also save the experimental time. Experimental results on four POLSAR datasets also demonstrate that our proposed methods are significantly effective.

Multi-Stream Convolutional Neural Network for SAR Automatic Target Recognition

Abstract: Despite the fact that automatic target recognition (ATR) in Synthetic aperture radar (SAR) images has been extensively researched due to its practical use in both military and civil applications, it remains an unsolved problem. The major challenges of ATR in SAR stem from severe data scarcity and great variation of SAR images. Recent work started to adopt convolutional neural networks (CNNs), which, however, remain unable to handle the aforementioned challenges due to their high dependency on large quantities of data. In this paper, we propose a novel deep convolutional learning architecture, called Multi-Stream CNN (MS-CNN), for ATR in SAR by leveraging SAR images from multiple views. Specifically, we deploy a multi-input architecture that fuses information from multiple views of the same target in different aspects; therefore, the elaborated multi-view design of MS-CNN enables it to make full use of limited SAR image data to improve recognition performance. We design a Fourier feature fusion framework derived from kernel approximation based on random Fourier features which allows us to unravel the highly nonlinear relationship between images and classes. More importantly, MS-CNN is qualified with the desired characteristic of easy and quick manoeuvrability in real SAR ATR scenarios, because it only needs to acquire real-time GPS information from airborne SAR to calculate aspect differences used for constructing testing samples. The effectiveness and generalization ability of MS-CNN have been demonstrated by extensive experiments under both the Standard Operating Condition (SOC) and Extended Operating Condition (EOC) on the MSTAR dataset. Experimental results have shown that our proposed MS-CNN can achieve high recognition rates and outperform other state-of-the-art ATR methods.