Search

Abstract The article proposes a real‐time apple picking method based on an improved YOLOv5. This method accurately recognizes different apple targets on fruit trees for robots and helps them adjust their position to avoid obstructions during fruit picking. Firstly, the original BottleneckCSP module in the YOLOv5 backbone network is enhanced to extract deeper features from images while maintaining lightweight. Secondly, the ECA module is embedded into the improved backbone network to better extract features of different apple targets. Lastly, the initial anchor frame size of the network is adjusted to avoid recognizing apples in distant planting rows. The results demonstrate that this improved model achieves high accuracy rates and recall rates for recognizing various types of apple picking methods with an average recognition time of 0.025s per image. Compared with other models tested on six types of apple picking methods, including the original YOLOv5 model as well as YOLOv3 and EfficientDet‐D0 algorithms, our improved model shows significant improvements in mAP by 1.95%, 17.6%, and 12.7% respectively. This method provides technical support for robots' picking hands to actively avoid obstructions caused by branches during fruit harvesting, effectively reducing apple loss.

Abstract Applying the sparse recovery (SR) technique to airborne radar space‐time adaptive processing (STAP) can greatly reduce the number of required training samples, which is advantageous in detecting targets in non‐homogeneous and non‐stationary clutter environments. However, the poor performance, the slow convergence speed or the high computational complexity of the traditional SR STAP algorithms limit their practical application. To tackle this problem, a novel efficient SR STAP algorithm is proposed. The newly proposed SR STAP algorithm utilises the log‐sum penalty to approximate the ℓ0−norm penalty, which exhibits improved convergence performance and clutter suppression performance compared to the traditional SR STAP algorithms. Besides, the proposed algorithm can ensure the convergence in each iteration by offering a closed‐form analytic solution. Furthermore, the result of the mathematical derivation demonstrates the essential equivalence between our method and the iterative reweighted ℓ2 method. By utilising this equivalence, two additional methods are proposed that incorporate the knowledge of the clutter Capon spectrum and the clutter spectrum of the iterative adaptive approach (IAA) as the components of weighted values, resulting in further performance improvement of the proposed algorithm. Finally, simulation results with both simulated data and Mountain‐Top data demonstrate the high effectiveness and performance of the proposed methods.

Abstract In the anti‐interference of broadband communication radio, interference cancellation performance is seriously affected by the uncertainty of the working frequency selection of the communication system and the direction of hostile interference. The closed‐form solution of the interference cancellation weight value and the signal‐to‐interference noise ratio after the cancellation are derived and simplified to clarify the factors affecting interference cancellation quality. Then, the efficient blind area criterion is established for the low computational complexity in view of cancellation quality, which combines the radio communication index and anti‐interference demand index. Based on this criterion, the minimum blind area model is proposed for robust performance. The experimental results in the auxiliary antenna spacing optimisation are consistent with the theoretical analysis, verifying the effectiveness of the proposed model in dealing with uncertain communication frequency and interference direction. It is very promising in communication interference cancellation.

Abstract The line of sight (LoS) probability is a key factor for the channel modeling of air‐to‐ground (A2G) communication. However, the existing LoS probability models do not account for the effects of airframe shadowing (AS) and building density, which can cause serious link obstruction and performance loss due to the six‐dimensional (6D) mobility and self‐body of unmanned aerial vehicle (UAV). In this article, a new LoS probability model is proposed that considers the AS and building density for different UAV heights. Adding to this, the AS is derived in terms of UAV framework and 6D mobility. Next, the machine learning (ML) based graph neural network (GNN) method is developed to learn the features and structure of the urban environment and predict the LoS probability. Then, the GNN model is trained and evaluated based on the ray tracing (RT) data to establish the relationship between model parameters and UAV heights under the building density and AS factors. The interpretation and explanation of the proposed GNN model and prediction are also discussed in this article. It is shown from the simulation analysis that the GNN model accurately captures the effects of AS, building height distributions, and UAV heights, with high accuracy compared to the baseline 3GPP, GCM and NYU models.

Abstract Space‐time adaptive processing (STAP) array has found extensive use in the global navigation satellite system for interference suppression. Nevertheless, the conventional space‐time array necessitates multiple antennas and time delay taps, resulting in a heavy burden on both hardware and computation resources. Thus, lowering the cost and complexity of STAP becomes a considerable problem demanding prompt solutions. This work designs a space‐time array reconfiguration scheme based on the joint selection strategy of the antenna and delay tap to address this problem. The reconstructed array can induce a high signal to interference plus noise power ratio (SINR) when using fewer antennas and delay taps. The space‐time correlation coefficient (STCC) is first presented to manifest the influence of reconfiguration on output SINR. Then, we formulate the space‐time array reconfiguration mathematical model, aiming to minimise STCC. Two reconfiguration methods are provided to solve the optimal array configuration of antennas and delay taps that can reach the minimum STCC and maximum SINR. The optimal antenna‐delay tap pairs are finally selected to comprise the reconstructed array. Substantial simulation experiments verify that the presented scheme and strategy are effective and reliable in space‐time array reconfiguration.

Abstract Rolling bearings are essential parts in machine equipment and detecting damage in the early stage is crucial for ensuring the safe production and machine life. However, it is difficult to extract weak fault features under strong background noise, discrete harmonic frequency interference and non‐stationary service conditions. This investigation proposes a hybrid fault diagnosis approach utilizing transient structure‐optimal variational mode decomposition (TS‐OVMD) and adaptive group sparse coding (AGSC) for addressing the formidable problem. According to the singular value structure between transient signal and the interference signal, this work investigates the singular value shrinkage (SVS) technique to adaptively obtain the independent components number. Then, we present a transient structure measure (TSM) to adaptively optimize the balance factor. This measure index systematically quantifies the typical characteristics of the bearing fault signal, which can maximize the fault information representation and effectively reduces the useful information loss caused by improper selection of VMD parameters. Finally, a sparse coding model called AGSC is furthermore designed to enhance the fault impulses readability and suppress residual noise based on the sparsity within group property and the TSM. The proposed approach is verified using experimental data and is found to be superiority comparison with the state‐of‐the‐art method.

Abstract This article proposes a cooperative jamming (CJ) architecture based on the coded signal of interest (SOI). The CJ is superimposed on the coded SOI directly at the transmitter, and the authorized user reconstructs the received signal and cancels the CJ based on the prior knowledge of the coding information, then different radiation patterns after cancellation are created for SOI and CJ, which effectively cancels CJ while retaining SOI. Eavesdropper, however, is unaware of the coding information and will cause significant loss to the SOI while cancelling the CJ, therefore unable to achieve eavesdropping on communication signals. Numerical results verify the feasibility of the proposed method.

Abstract Deep learning‐based‐image denoising methods have recently achieved excellent performance by learning non‐linear mapping in the spatial domain. However, these methods fail to address the noise without specific distribution because they only use features of the spatial domain. Meanwhile, existing methods that utilize features in the frequency domain fail to combine the detailed information of both domains properly for effective reconstruction and demonstrate poor generalization. Therefore, a novel adaptive fusion dual‐domain network (AFDN) is introduced for single image restoration. Different from deep learning‐based methods, which operate on the spatial or dual‐domains in a certain order, the proposed AFDN combine the spatial‐domain image and corresponding frequency‐domain image as the input and use the interlacing dual‐domain module with flexible adaptability to learn the relationship between spatial and frequency domains. In experimental results, the AFDN is compared with several state‐of‐the‐art restoration methods. Quantitative results showed that the AFDN achieves enhanced effects and high index values. The code of this paper will be released at https://github.com/jzw0707/AFDN

Abstract This paper presents an investigation into the theoretical performance of diffusion adaptive networks within a distributed beam coordination application in a cell‐free communication network that utilizes coarsely quantized data. This analysis assumed data pulses with 1, 2, and 3 bits, and based on this assumption, simulations and theoretical analyses of distributed beamforming are conducted. The consistency between the simulation results and theoretical findings validated the accuracy of the calculations. Additionally, the distributed quantization aware least mean square (DQA‐LMS) algorithm was introduced for distributed beam coordination, demonstrating its superior performance in scenarios involving coarsely quantized data. Convergence conditions were provided and the results through various simulations were presented, including the array network mean square deviation (MSD), mean square error (MSE), and beamforming diagrams.

Abstract Due to the controllable degrees of freedom of frequency diverse array in the distance dimension, the frequency diverse multiple‐input multiple‐output (FDA‐MIMO) radar can easily deal with mainlobe interference, which is often difficult to solve in traditional phased arrays. Nevertheless, the performance of FDA‐MIMO radar will suffer from degradation in the scenario where spectrum interferences and signal‐dependent interferences coexist. In order to solve this problem, a novel FDA‐MIMO radar framework via antenna switching is proposed. Based on this framework, an effective method for jointly optimising antenna selection and beamforming against spectrum interferences and signal‐dependent interferences is developed. The resulting optimization problem is nonconvex and NP‐hard owing to the integer constraints caused by antenna selection. By relaxing the integer constraints, the original problem can be transformed into a convex optimization form and an iterative reweighting strategy is used to force the obtained solution to satisfy the integer constraints. Simulation examples show that the proposed algorithm has better performance in output signal‐to‐interference‐plus‐noise ratio (SINR) and beampattern radiation performance than the existing competitive methods.

Abstract This paper proposes a new secure distributed estimation for cyber‐physical systems against adversarial attacks with noisy input. To mitigate the effect of attacks, a novel distributed attack detection based on a reliable reference estimation obtained by temporal convex combination is proposed. Furthermore, for more effective and robust performance, an adaptation rule to adjust convex combination weights is presented, in which the generalized correntropy method with nonlinear loss function and stochastic gradient descent are utilized. Besides, to eliminate input noise, a bias‐compensation method in local adaptation of the secure distributed estimation is proposed. Simulations show superior dynamic and real‐time adaptability of the proposed algorithm under complex attacking scenarios.

Abstract The jamming effect on radar is dominated by the design of the jamming waveform directly. Traditional jamming waveforms are generated using the template‐based method and are not environmentally resilient. Without the precise radar centre frequency support, using the direct guidance of the radar signal power spectrum density (PSD), a novel jamming waveform design method with an online unsupervised generative learning framework is proposed. The proposed framework consists of two modules: the waveform online generation module and the waveform unsupervised optimisation module. For the waveform online generation module, the well‐designed neural network (NN) is used to generate the jamming waveform. While for the waveform unsupervised iterative optimisation module, two loss functions are developed as two learning tasks to guide back‐propagation online. Given transmit power, the jamming PSD is designed adaptively to focus on the radar band and fit the radar PSD well, which improves the receiver in‐band jamming‐to‐signal ratio (JSR). It is worth noting that NN is end‐to‐end trained online without extensive prior training data. The framework can generate jamming waveforms within reasonable physical parameter ranges, and numerical experiments have shown that the designed jamming waveform performed better jamming effects on radar target detection in comparison with the traditional one.

Abstract Non‐sidelooking airborne radar encounters significant non‐stationary and heterogeneous clutter environments, resulting in a severe shortage of samples. Sparse recovery‐based space‐time adaptive processing (SR‐STAP) methods can achieve good clutter suppression performance with limited samples. Nonetheless, grid‐based SR‐STAP algorithms encounter off‐grid effects in non‐sidelooking arrays, which can severely degrade the clutter suppression performance. In this study, the authors propose a novel gridless SR‐STAP method in the continuous spatial‐temporal domain to address the issue of off‐grid effects. Inspired by the fact that sparse Bayesian learning (SBL) framework implicitly performs a structured covariance matrix estimation, the authors reparameterise its cost function to directly estimate the block‐Toeplitz structured matrix from the measurements in a gridless manner. Since the proposed cost function is non‐convex, we utilise a majorisation‐minimisation‐based iterative procedure to estimate the clutter covariance matrix. Finally, using the standard concept of semidefinite programming, the authors derive a convex gridless implementation of the SBL cost function for uniformly sampled radar systems. Extensive simulation experiments demonstrate the exceptional clutter suppression and target detection performance of the proposed algorithm.

Abstract This article deals with the source localisation problem when the receiver is moving over time, and the sources are sparse across the area. In addition, only one of the sources is active at each time instant, means that the sources are pulse transmitters. The problem was first formulated such as a dictionary learning problem for sparse representation. Then, an alternating minimisation approach to estimate the position of the sources and their signals iteratively is proposed. Simulation results in different scenarios demonstrate the effectiveness of the proposed method relative to state‐of‐the‐art methods.

Abstract The trajectory prediction of the hypersonic glide vehicle (HGV) can provide hit point information for early warning systems, which is of great significance for near‐space defence operations. However, the heavy‐tailed noise caused by abnormal environmental disturbance and abrupt changes in vehicle trajectory seriously affects the accuracy of HGV trajectory prediction. To solve the problem of trajectory prediction for HGV under heavy‐tailed noise, we propose an adaptive multivariate Student's t‐process regression method called aEM‐MVTPR. Firstly, a heavy‐tailed noise model is developed using the Student's t‐distribution. Secondly, a multivariate Student's t‐process regression (MVTPR) for HGV trajectory prediction is derived, and the method can learn the HGV trajectory time series features and mine the correlations among the trajectory variables. Finally, to further improve the method's robustness, we use the accelerated expectation maximisation algorithm and Pearson correlation analysis to adaptively estimate and adjust the initial values of the MVTPR parameters. The simulation experimental results show that the proposed method has more accurate predicted values of trajectory variables than multivariate Gaussian process regression (MVGPR) under heavy‐tailed noise conditions. In addition, the ability of the aEM‐MVTPR to adaptively adjust the parameter makes it more robust than the MVTPR under different noise environments.

Abstract Visibility graph methods allow time series to mine non‐Euclidean spatial features of sequences by using graph neural network algorithms. Unlike the traditional fixed‐rule‐based univariate time series visibility graph methods, a symmetric adaptive visibility graph method is proposed using orthogonal signals, a method applicable to in‐phase and quadrature (I/Q) orthogonal signals for adaptive graph mapping for radio modulated signals in automatic modulation classification tasks. The method directly models the intra‐channel and inter‐channel graph relations of I/Q signals using two different types of convolutional kernels. It captures non‐Euclidean spatial feature information of I/Q signals using a graph neural network combining graph sampling aggregation and graph differentiable pooling as a feature extractor. Extensive experimental results on two benchmark datasets and a simulated dataset containing channel fading show that the proposed Quadrature Signal Symmetric Adaptive Visibility Graph (QSSAVG) method in this paper outperforms the benchmark method in terms of classification accuracy and is also more robust against channel fading and noise variations.

Abstract Light‐field (LF) images offer the potential to improve feature capture in live scenes from multiple perspectives, and also generate additional normal vectors for performing super‐resolution (SR) image processing. With the benefit of machine learning, established AI‐based deep CNN models for LF image SR often individualize the models for various resolutions. However, the rigidity of these approaches for actual LF applications stems from the considerable diversity in angular resolution among LF instruments. Therefore, an advanced neural network proposal is required to utilize a CNN‐based model for super‐resolving LF images with different resolutions obtained from provided features. In this work, a preprocessing to calculate the depth channel from given LF information is first presented, and then a multiple‐decouple and fusion module is introduced to integrate the VGGreNet for the LF image SR, which collects global‐to‐local information according to the CNN kernel size and dynamically constructs each view through a global view module. Besides, the generated features are transformed to a uniform space to perform final fusion, achieving global alignment for precise extraction of angular information. Experimental results show that the proposed method can handle benchmark LF datasets with various angular and different resolutions, reporting the effectiveness and potential performance of the method.

Abstract Space‐time adaptive processing (STAP) struggles to effectively suppress clutter in the heterogeneous clutter environment due to the lack of training samples. In order to enhance clutter suppression performance of STAP, a subspace‐weighted mixed‐norm minimisation approach is given. First, a roughly estimated clutter subspace is obtained using the subspace augment (SA) approach. The weight vector is then designed using the association between the dictionary matrix and the noise subspace, allowing the algorithm to penalise sparse coefficients democratically. Finally, in order to solve the subspace‐weighted mixed‐norm minimisation problem, we derive a fast algorithm based on the alternating direction multiplier method (ADMM) framework. The proposed algorithm does not require iteratively updating the weight vector in contrast to the iterative re‐weighted l1 ${l}_{1}$ (IRL1) algorithm. The simulation results demonstrate the effectiveness of the proposed algorithm in terms of computational efficiency and clutter suppression performance.

Abstract In this paper, a novel color image encryption scheme based on fractional discrete cosine transform (FrDCT) and discrete wavelet transform (DWT) with interplane arrangements in the discrete orthonormal Stockwell transform (DOST) domain is presented. Color images are encrypted using the keys of FrDCT and coefficients of interplane arrangement of DWT sub‐bands. These keys are applied on three independent planes, that is, R, G, and B. To ensure the correct decryption of the encrypted image, every key need to be clearly understood in the correct sequence and its exact values. FrDCT has the advantageous feature of storing real‐valued coefficients, and the energy of the image can be represented with very few coefficients. Therefore, transmitting and storing the encrypted image becomes easy. FrDCT may be considered a real‐valued cosine version of fractional Fourier transform (FrFT), which is used successfully for storage and transmission due to its property of compacting energy distribution. Furthermore, for more robustness, interplane arrangements are explored. The concept of interplane arrangements has not been substantially reported in the existing literature. The comparison of the technique with other similar encryption techniques available in the literature has been carried out by performing different experiments to examine the efficacy of the proposed method. It is found that the proposed method outperforms other existing techniques.

Abstract Statistical features are commonly used in radar target recognition tasks using radar cross section as the data source. The traditional statistical feature extraction way may not make better use of the data information, making the target classification results slightly worse. Here, a new sliding window segmentation method is proposed for extracting statistical features to improve the accuracy of the radar target classification task and investigate the separability of the statistical features extracted after sliding window segmentation. The results show that the separability of the statistical features extracted by our method is better than other traditional ways. The classification accuracy of using BP neural network, SVM and KNN classification algorithms is improved by about 10% compared with the traditional way.

The architecture and processing algorithm of mainlobe interference suppression method is described for nulling the signal from mainlobe electronic jammer and multiple sidelobe electronic jammers while maintaining monopulse angle estimation accuracy on the target. The existing mainlobe interference suppression method is based on monopulse Wiener filtering. However, the Wiener filtering is more suitable to remove static interference. Therefore, the authors present the dynamic mainlobe interference suppression method with Gray Model Kalman filter to resolve this problem. The simulations verify the proposed method.

Cognitive passive radar (CPR) system is amenable to electromagnetic environment cooperating and optimal strategy making tasks on account of illumination source availability and waveform evaluation. In fact, intelligence will provide performance gain from multi-dimension optimisation. Therefore, we propose a multi-standard multi-scheme compatible wideband CPR framework which selects the most suitable waveform as the passive radar illumination source. First, we develop a software defined radio platform to realise wideband and flexible signal processing, which provides the feasibility of cognition. Then in order to recognise the electromagnetic waveforms, we present a deep learning framework which can be integrated in CPR system as a classifier using convolutional neural networks (CNNs) and long short-term memory (LSTM). Via the proposed innovative passive perception-action cycle, the CPR prototype system demonstrates the real-time, high accuracy modulation recognition ability and the adaptive signal processing ability. Based on our experiments, it can be demonstrated that our CPR prototype system integrated with deep learning is effective and promising.

Cloud computing offers significant research and economic benefits to healthcare organisations. Cloud services provide a safe place for storing and managing large amounts of such sensitive data. Under conventional flow of gene information, gene sequence laboratories send out raw and inferred information via Internet to several sequence libraries. DNA sequencing storage costs will be minimised by use of cloud service. In this study, the authors put forward a novel genomic informatics system using Amazon Cloud Services, where genomic sequence information is stored and accessed for processing. True identification of exon regions in a DNA sequence is a key task in bioinformatics, which helps in disease identification and design drugs. Three base periodicity property of exons forms the basis of all exon identification techniques. Adaptive signal processing techniques found to be promising in comparison with several other methods. Several adaptive exon predictors (AEPs) are developed using variable normalised least mean square and its maximum normalised variants to reduce computational complexity. Finally, performance evaluation of various AEPs is done based on measures such as sensitivity, specificity and precision using various standard genomic datasets taken from National Center for Biotechnology Information genomic sequence database.

Cloud computing offers significant research and economic benefits to healthcare organisations. Cloud services provide a safe place for storing and managing large amounts of such sensitive data. Under conventional flow of gene information, gene sequence laboratories send out raw and inferred information via Internet to several sequence libraries. DNA sequencing storage costs will be minimised by use of cloud service. In this study, the authors put forward a novel genomic informatics system using Amazon Cloud Services, where genomic sequence information is stored and accessed for processing. True identification of exon regions in a DNA sequence is a key task in bioinformatics, which helps in disease identification and design drugs. Three base periodicity property of exons forms the basis of all exon identification techniques. Adaptive signal processing techniques found to be promising in comparison with several other methods. Several adaptive exon predictors (AEPs) are developed using variable normalised least mean square and its maximum normalised variants to reduce computational complexity. Finally, performance evaluation of various AEPs is done based on measures such as sensitivity, specificity and precision using various standard genomic datasets taken from National Center for Biotechnology Information genomic sequence database.