特征增强的Sparse Transformer目标跟踪算法

针对Transformer的自注意力机制计算量大、容易被背景分心,导致有效信息抓取不足,从而降低跟踪性能的问题,提出特征增强的Sparse Transformer目标跟踪算法。基于孪生网络骨干进行特征提取;特征增强模块利用多尺度特征图生成的上下文信息,增强目标局部特征;利用Sparse Transformer的最相关特性生成目标聚焦特征,并嵌入位置编码提升跟踪定位的精度。提出的跟踪模型以端到端的方式进行训练,在OTB100,VOT2018和LaSOT等5个数据集上进行了大量实验,实验结果表明所提算法取得了较好的跟踪性能,实时跟踪速度为34帧/s。

Radar fast long-time coherent integration via TR-SKT and robust sparse FRFT

Long-time coherent integration(LTCI)is an effective way for radar maneuvering target detection,but it faces the problem of a large number of search parameters and large amount of calculation.Realizing the simultaneous compensation of the range and Doppler migrations in complex clutter back-ground,and at the same time improving the calculation efficiency has become an urgent problem to be solved.The sparse transformation theory is introduced to LTCI in this paper,and a non-parametric searching sparse LTCI(SLTCI)based maneuvering target detection method is proposed.This method performs time reversal(TR)and second-order Keystone transform(SKT)in the range frequency&slow-time data to complete high-order range walk compensation,and achieves the coherent integra-tion of maneuvering target across range and Doppler units via the robust sparse fractional Fourier transform(RSFRFT).It can compensate for the nonlinear range migration caused by high-order motion.S-band and X-band radar data measured in sea clutter background are used to verify the detection performance of the proposed method,which can achieve better detection performance of maneuvering targets with less computational burden compared with several popular integration methods.

Reconstruction method of irregular seismic data with adaptive thresholds based on different sparse transform bases

Oil and gas seismic exploration have to adopt irregular seismic acquisition due to the increasingly complex exploration conditions to adapt to complex geological conditions and environments.However,the irregular seismic acquisition is accompanied by the lack of acquisition data,which requires high-precision regularization.The sparse signal feature in the transform domain in compressed sensing theory is used in this paper to recover the missing signal,involving sparse transform base optimization and threshold modeling.First,this paper analyzes and compares the effects of six sparse transformation bases on the reconstruction accuracy and efficiency of irregular seismic data and establishes the quantitative relationship between sparse transformation and reconstruction accuracy and efficiency.Second,an adaptive threshold modeling method based on sparse coefficient is provided to improve the reconstruction accuracy.Test results show that the method has good adaptability to different seismic data and sparse transform bases.The f-x domain reconstruction method of effective frequency samples is studied to address the problem of low computational efficiency.The parallel computing strategy of curvelet transform combined with OpenMP is further proposed,which substantially improves the computational efficiency under the premise of ensuring the reconstruction accuracy.Finally,the actual acquisition data are used to verify the proposed method.The results indicate that the proposed method strategy can solve the regularization problem of irregular seismic data in production and improve the imaging quality of the target layer economically and efficiently.

An Efficient Radar Detection Method of Maneuvering Small Targets

Detection of maneuvering small targets has always been an important yet challenging task for radar signal processing.One primary reason is that target variable motions within coherent processing interval generate energy migrations across multiple resolution bins,which severely deteriorate the parameter estimation performance.A coarse-to-fine strategy for the detection of maneuvering small targets is proposed.Integration of small points segmented coherently is performed first,and then an optimal inter-segment integration is utilized to derive the coarse estimation of the chirp rate.Sparse fractional Fourier transform(FrFT)is then employed to refine the coarse estimation at a significantly reduced computational complexity.Simulation results verify the proposed scheme that achieves an efficient and reliable maneuvering target detection with-16dB input signal-to-noise ratio(SNR),while requires no exact a priori knowledge on the motion parameters.

Fast GNSS Acquisition Algorithm Based on SFFT with High Noise Immunity

The Global Navigation Satellite System(GNSS)has been widely used in various fields.To achieve positioning,the receiver must first lock the satellite signal.This is a complicated and expensive process that consumes a lot of resources of the receiver.For this reason,this paper proposes a new fast acquisition algorithm with High Signal-tonoise ratio(SNR)performance based on sparse fast Fourier transform(HSFFT).The algorithm first replaces the IFFT process of the traditional parallel code phase capture algorithm with inverse sparse fast Fourier transform(ISFFT)with better computing performance,and then uses linear search combined with code phase discrimination to replace the positioning loop and the estimation loop with poor noise immunity in ISFFT.Theoretical analysis and simulation results show that,compared with the existing SFFT parallel code phase capture algorithm,the calculation amount of this algorithm is reduced by 19%,and the SNR performance is improved by about 5dB.Compared with the classic FFT parallel code phase capture algorithm,the calculation amount of the algorithm in this paper is reduced by 43%,and when the capture probability is greater than 95%,the SNR performance of the two is approximately the same.

Self-Sparse Generative Adversarial Networks

Generative adversarial networks(GANs)are an unsupervised generative model that learns data distribution through adversarial training.However,recent experiments indicated that GANs are difficult to train due to the requirement of optimization in the high dimensional parameter space and the zero gradient problem.In this work,we propose a self-sparse generative adversarial network(Self-Sparse GAN)that reduces the parameter space and alleviates the zero gradient problem.In the Self-Sparse GAN,we design a self-adaptive sparse transform module(SASTM)comprising the sparsity decomposition and feature-map recombination,which can be applied on multi-channel feature maps to obtain sparse feature maps.The key idea of Self-Sparse GAN is to add the SASTM following every deconvolution layer in the generator,which can adaptively reduce the parameter space by utilizing the sparsity in multi-channel feature maps.We theoretically prove that the SASTM can not only reduce the search space of the convolution kernel weight of the generator but also alleviate the zero gradient problem by maintaining meaningful features in the batch normalization layer and driving the weight of deconvolution layers away from being negative.The experimental results show that our method achieves the best Fréchet inception distance(FID)scores for image generation compared with Wasserstein GAN with gradient penalty(WGAN-GP)on MNIST,Fashion-MNIST,CIFAR-10,STL-10,mini-ImageNet,CELEBA-HQ,and LSUN bedrooms datasets,and the relative decrease of FID is 4.76%-21.84%.Meanwhile,an architectural sketch dataset(Sketch)is also used to validate the superiority of the proposed method.