This paper presents a new semantic sparse recoding method to generate more descriptive and robust representation of visual content for image applications. Although the visual bag-of-words (BOW) representation has been reported to achieve promising results in different image applications, its visual codebook is completely learnt from low-level visual features using quantization techniques and thus the so-called semantic gap remains unbridgeable. To handle such challenging issue, we utilize the annotations (predicted by algorithms or shared by users) of all the images to improve the original visual BOW representation. This is further formulated as a sparse coding problem so that the noise issue induced by the inaccurate quantization of visual features can also be handled to some extent. By developing an efficient sparse coding algorithm, we successfully generate a new visual BOW representation for image applications. Since such sparse coding has actually incorporated the high-level semantic information into the original visual codebook, we thus consider it as semantic sparse recoding of the visual content. Finally, we apply our semantic sparse recoding method to automatic image annotation and social image classification. The experimental results on several benchmark datasets show the promising performance of our semantic sparse recoding method in these two image applications.

Semantic sparse recoding of visual content for image applications.

In this paper, convolutional neural networks (CNN)-based ship target recognition is studied by exploiting the targets' high resolution range profiles (HRRPs). Contrary to conventional procedures employing hand-crafted features, by designing an appropriate CNN scheme, features are learned automatically in order through convolutional layers and, recognition of military and civilian ship targets is performed. In order to simulate the targets' scatterings accurately, their realistic computer-aided design (CAD) models are considered. Additionally, scattering characteristics of the targets are taken into account for a variety of azimuthal and elevation aspects. Promising simulation results exhibit that CNN-based schemes would provide easiness and enhanced performance in ship target recognition area due to their self-feature learning nature.

Convolutional neural networks-based ship target recognition using high resolution range profiles

This paper presents a novel L1-norm semi-supervised learning algorithm for robust image analysis by giving new L1-norm formulation of Laplacian regularization which is the key step of graph-based semi-supervised learning. Since our L1-norm Laplacian regularization is defined directly over the eigenvectors of the normalized Laplacian matrix, we successfully formulate semi-supervised learning as an L1-norm linear reconstruction problem which can be effectively solved with sparse coding. By working with only a small subset of eigenvectors, we further develop a fast sparse coding algorithm for our L1-norm semi-supervised learning. Due to the sparsity induced by sparse coding, the proposed algorithm can deal with the noise in the data to some extent and thus has important applications to robust image analysis, such as noise-robust image classification and noise reduction for visual and textual bag-of-words (BOW) models. In particular, this paper is the first attempt to obtain robust image representation by sparse co-refinement of visual and textual BOW models. The experimental results have shown the promising performance of the proposed algorithm.

Robust Image Analysis by L1-Norm Semi-supervised Learning

The invention relates to a radar imaging method of an electrically large size target in the ocean clutter environment. The radar imaging method is that the radar imaging of the electrically large size target in the ocean clutter environment is achieved by establishing a surface-box complex target scattering model in a multi-interference environment and adopting electromagnetic simulating calculation through a numerical and analytical hybrid algorithm and a rapid back projection imaging mode. The electrically large size in the radar imaging method means that the ratio of the physical dimension to the wavelength of the target is larger than ten. The working frequency range of a radar is generally in a high frequency area, for example, the L waveband frequency ranges from 1GHz to 2 GHz, and the S waveband frequency ranges from 2GHz to 4 GHz. Some military targets, such as fighters, invisible planes and aircraft carriers in the L waveband and the S waveband belong to the electrically large size. According to the radar imaging method, the radar imaging of the electrically large size target in the ocean clutter environment can be achieved by means of geometric modeling of complex targets, the electromagnetic simulating calculation, the rapid back projection imaging algorithm and Kaiser window edge filtering.

Radar imaging method of electrically large size target in ocean clutter environment

The invention relates to a method for estimating the low-altitude electromagnetic wave propagation characteristic on the basis of sea clutter and belongs to the technical field of real-time estimation of the marine environment characteristic of the electromagnetic wave propagation characteristic. The method is mainly used for solving the problems of estimating a marine atmosphere refractive index section parameter and evaluating the electromagnetic wave propagation characteristic. Different from a method for estimating the radar detection efficacy on the basis of a sea clutter power, the method of the invention is used for obtaining an atmosphere refractive index section drawing and an electromagnetic wave propagation attenuation drawing mainly by inverting an actual atmosphere refractive index section and an electromagnetic wave propagation attenuation quantity, solving the atmosphere refractive index section parameter and then checking estimated values on the basis of evaluation of an actual measurement sea clutter power spectrum and the solution and comparison of an electromagnetic wave propagation diffusion power spectrum by presetting parameters of an atmosphere refractive index section model. The method of the invention is simple for engineering implementation and has the advantages of good military concealment, small prediction error and low cost; in addition, the equipment can be embedded and loaded; and the like. The method is mainly applied to measuring and estimating the atmosphere environment and the electromagnetic wave propagation attenuation variation in the military and marine monitoring.

Method for estimating low-altitude electromagnetic wave propagation characteristic on basis of sea clutter

Radar cross-section layered calculation method of ship target within atmospheric duct range

For the linear frequency-modulated continuous wave radar, the coupling of fast time and slow time in the signal phase is occurred for the moving target detection, which results in migration through range cell in multi-cycle signal. A new processing method for moving target detection based on Keystone transform is proposed to solve above problem. Through the characteristics analysis of the received multi-cycle signal and utilizing the decoupling property of Keystone transform, a transform expression in the moving target detection processing, is derived which can be used for the range profiles alignment. Finally realization and performance of the proposed method is discussed. Compared with the direct motion compensation method, the efficiency of the proposed method and the relevant conclusion are verified by the simulation results.

MTD processing based on Keystone transform for LFMCW radar

The invention discloses an underwater acoustic measurement method for scattering properties of a ship target broadband radar, which is characterized by comprising the following steps of: firstly, inversely arranging a scaled model of a measured ship target in an anechoic tank, regulating basic conditions of a measurement device and a modeling test, simulating irregular sea waves at an interface between the scaled model of the ship target and water, amplifying a linear frequency demodulation signal generated by a signal source and radiating the amplified frequency demodulation signal in a sonic signal manner, amplifying and filtering an echo output signal reflected by the scaled model and then storing, then acquiring a plurality of groups of echo signal data on each rotating angle and then storing; and finally, processing the echo signals of the scaled model of the measured ship target on different rotating angles by adopting a pulse compression technology to obtain one-dimensional distance image data of the measured ship target. The invention overcomes the defects of rigorous requirements on measurement equipment and environment and incapability of measuring sea-surface target characteristics in the prior art, and has the advantages of flexibility in the measurement method, low cost, easiness in realization and the like.

Underwater acoustic measurement method for scattering properties of ship target broadband radar

This paper presented here introduces a technique for automatic recognition of ship targets from high range resolution range profiles The technique incorporates Fourier-Mellin transform (FMT) and support vector machine (SVM) to identify ship targets According to the shift invariance of the Fourier transform and scale invariance of the Mellin transform, the high resolution range profiles are processed using FMT to form the feature vectors Within the recognizer, the multi-class SVM is used based on binary tree, which can reduce the number of SVM classifiers and repetitive training samples effectively The recognition results from simulation data and measurement data of ship target range profiles prove the validity and robustness of the method proposed

Ship target recognition using high resolution range profiles based on FMT and SVM

In the real-world applications, the detection performance of the target is unknown before the sensor captures the target. In this paper, a sensor management method for target search is presented with unknown detection performance. Firstly, a search model based on maximal information gain criterion is proposed, in which the information gain of each cell is predicted before the sensor detection, and the observation is made in the cell with the maximal information gain. Secondly, beta distribution is used to model the unknown detection performance and its posterior probability density with the sensor observation is derived based on Bayesian theory. Compared with the conventional method, simulation results show that the proposed search strategy can distinctly improve the search performance and the sensor can capture target with much less time resources.

Jianbin Lu

Papers

Radar cross-section layered calculation method of ship target within atmospheric duct range

MTD processing based on Keystone transform for LFMCW radar

Underwater acoustic measurement method for scattering properties of ship target broadband radar

Ship target recognition using high resolution range profiles based on FMT and SVM

Sensor Management for Target Search with Unknown Detection Performance