A METHOD OF AIRCRAFT TARGET RECOGNITION BASED ON LLE AND HMM

Xinde Li,Yilin Dong,Chaomin Luo,Yanliang Tan

DOI: https://doi.org/10.2316/journal.206.2017.2.206-4819

2017-01-01

Abstract:Aiming to the problem of low recognition rate for aircraft targetrecognition caused by the change of postures, an aircraft targetrecognition based on locally linear embedding (LLE) and hiddenMarkov model (HMM) is proposed in this paper.

What problem does this paper attempt to address?

Automatic Aircraft Recognition using DSmT and HMM

Xinde Li,Jin-dong Pan,Jean Dezert

2014-01-01

Abstract:In this paper we propose a new method for solving the Automatic Aircraft Recognition (AAR) problem from a sequence of images of an unknown observed aircraft. Our method exploits the knowledge extracted from a training image data set (a set of binary images of different aircrafts observed under three different poses) with the fusion of information of multiple features drawn from the image sequence using Dezert-Smarandache Theory (DSmT) coupled with Hidden Markov Models (HMM). The first step of the method consists for each image of the observed aircraft to compute both Hu's moment invariants (the first features vector) and the partial singular values of the outline of the aircraft (the second features vector). In the second step, we use a probabilistic neural network (PNN) based on the training image dataset to construct the conditional basic belief assignments (BBA's) of the unknown aircraft type within the set of a predefined possible target types given the features vectors and pose condition. The BBA's are then combined altogether by the Proportional Conflict Redistribution rule #5 (PCR5) of DSmT to get a global BBA about the target type under a given pose hypothesis. These sequential BBA's give initial recognition results that feed a HMM-based classifier for automatically recognizing the aircraft in a multiple poses context. The last part of this paper shows the effectiveness of this new Sequential Multiple-Features Automatic Target Recognition (SMF-ATR) method with realistic simulation results. This method is compliant with realtime processing requirement for advanced AAR systems.
Target Location Of Loitering Munitions Based On Image Matching

Zhifei Wang,Hua Wang

DOI: https://doi.org/10.1109/ICIEA.2011.5975657

2011-01-01

Abstract:This paper presented a method for high accuracy ground target localization by using Loitering Munitions (LM) equipped with a video camera sensor. Firstly, based on tray tracing ground target geo-location was used for comparison against our approach, and the calculation from pixel to world coordinates can be done. Then, in order to remove small details hough transform was used to image alignment and a median filter was applied which are visible from the sensed image but not visible from the reference image. Finally, based on bifurcation extraction a novel edge detection method and an image matching algorithm were proposed. This method did not require accurate knowledge of the aircraft position, attitude and high performance sensors. Therefore, it is especially suitable for LM which did not have capability to carry accurate sensors due to their limited play weight and power resources. The results of simulation experiments and theory analyzing demonstrate that high accuracy ground target localization is reached also using low performance sensors, and achieved the real time need. Additional, comparing with the matching algorithm based on normal cross correlation, the proposed algorithm has lower computational burden and faster processing speed, which can meet the real-time requirements for scene matching.
High Accuracy Ground Target Location Using Loitering Munitions Platforms

Wang Zhifei,Wang Hua,Han Jing

DOI: https://doi.org/10.1117/12.900519

2011-01-01

Abstract:Precise ground target localization is an interesting problem and relevant not only for military but also for civilian applications, and this is expected to be an emerging field with many potential applications. Ground Target Location Using Loitering Munitions (LM) requires estimation of aircraft position and attitude to a high degree of accuracy, and data derived by processing sensor images might be useful for supplementing other navigation sensor information and increasing the reliability and accuracy of navigation estimates during this flight phase. This paper presents a method for high accuracy ground target localization using Loitering Munitions (LM) equipped with a video camera sensor. The proposed method is based on a satellite or aerial image matching technique. In order to acquire the target position of ground intelligently and rapidly and to improve the localization accuracy estimating the target position jointly with the systematic LM and camera attitude measurement errors, several techniques have been proposed. Firstly, ground target geo-location based on tray tracing was used for comparison against our approach. By proposed methods the calculation from pixel to world coordinates can be done. Then Hough transform was used to image alignment and a median filter was applied for removing small details which are visible from the sensed image but not visible from the reference image. Finally, A novel edge detection method and an image matching algorithm based on bifurcation extraction were proposed. This method did not require accurate knowledge of the aircraft position and attitude and high performance sensors, therefore it is especially suitable for LM which did not have capability to carry accurate sensors due to their limited play weight and power resources. The results of simulation experiments and theory analyzing demonstrate that high accuracy ground target localization is reached with low performance sensors, and achieve timely. The method is used in reconnaissance and surveillance missions, or applicable in any other environment with a relevantly structured clutter.
Automatic method for aircraft target recognition in remote sensing images based on improved active shape model

Shoufei GAN,Hao SUN

DOI: https://doi.org/10.11835/j.issn.1000-582X.2014.09.011

2014-01-01

Abstract:We propose an estimation method for aircraft target direction based on histograms of oriented gradients (HOG),and then use the improved active shape model (ASM)to model the deformation between the different types of targets.Finally,we use kernel density estimation method (KDE)global statistical shape constraint to obtain the target to achieve the target recognition,and design a semi-automatic image feature point detection strategy for aircrafts, which improves the efficiency of training samples for calibration feature points.Recognition experiments on aircraft remote sensing images show the proposed method can better recognize aircraft targets than the existing methods.
Aircraft Type Recognition Based on Target Track

Honglan Huang,Jincai Huang,Yanghe Feng,Zhong Liu,Tao Wang,Li Chen,Yuzhen Zhou

DOI: https://doi.org/10.1088/1742-6596/1061/1/012015

2018-01-01

Journal of Physics Conference Series

Abstract:The target recognition plays an important role in air traffic management, and the research on automatic aircraft recognition is still in the exploratory stage. Since aircrafts move at high speeds in complex background, it is still a challenging task to quickly extract valid features from small amounts of data for aircraft type recognition. The machine learning is capable of capturing discriminating information and can well identify potential patterns from the dataset. Thus, we employ a novel classification model based on machine learning with the use of some effective motion features extracted from aircraft flight track information as the input of the model such as maximum speed, cruise speed, maximum acceleration, maximum climb rate. We performed experiments on datasets collected by multiple sensors and the results demonstrated the effectiveness of proposed model in significantly improving the aircraft type recognition accuracy, especially for military aircrafts. Besides, the model avoids the inherent deficiencies of graphics and image processing, and can well meet the needs of the air traffic management.
Aircraft Recognition Based on Improved Iterative Threshold Selection and Skeleton Zernike Moment

Yiliang Zeng,Jinhui Lan,Chuanzhao Han,Kewei Huang,Jiehui Li,Xuefei Shi

DOI: https://doi.org/10.1016/j.ijleo.2014.01.135

IF: 3.1

2014-01-01

Optik

Abstract:Using the visible optics images to identify targets is an important part in the development of remote sensing technology. In this paper, a new aircraft recognition method based on the improved iterative threshold selection and the skeleton Zernike moment is presented. The method segment aircraft targets under complex background using iterative threshold selection with between-class distance and scatter, and calculate the skeleton Zernike moment for the aircraft target recognition using template matching method. The experimental results show that the new method can effectively achieve the target segmentation under complex backgrounds, and provide a satisfactory performance both in recognition rate and recognition speed.
Automatic Target Recognition Method Based on Sequential Images

Jin HUANG,Yan LIANG,Yong-mei CHENG,Quan PAN,Jin-wen HU

DOI: https://doi.org/10.3321/j.issn:1000-6893.2006.01.017

2006-01-01

Abstract:It is much difficult to recognize a 3D target just based on a single 2D target image because of the multivocal information.In this paper,an automatic target recognition method based on sequential 2D images is proposed.Firstly,the modified Hu invariant moments are used as the invariant characteristic vectors,which are further inputed to a back-propagation neural network(BPNN) classifier.Then the BPNN classifier gives the primary recognition result,which is combined with the training error to achieve the basic belief assignment(BBA).Finally,a revised Dempster-Shafer(DS) reasoning method named absorption method,which can deal with high conflicting evidences,is applied to implement the final reasoning decision.The simulation based on multiple aircraft images with various attitudes demonstrates that the proposed method can recognize the aircrafts quickly and accurately.Besides this,this method has strong robustness to a priori parameter and the attitude variety of aircraft images.
A Novel HRRP Target Recognition Method Based on LSTM and HMM Decision-making.

Jun Tu,Teng Huang,Xusong Liu,Fei Gao,Erfu Yang

DOI: https://doi.org/10.23919/iconac.2019.8895040

2019-01-01

Abstract:Feature learning is a key step of target recognition for high-resolution range profile (HRRP). In traditional methods, Hidden Markov Model (HMM) can learn the features of HRRP target aspect information. However, the contextual correlations of HRRPs are ignored due to the independence assumptions in HMM, which brings many limitations to feature learning and weakens the generalization performance. On the contrary, the Long Short-Term Memory (LSTM) network can learn the contextual correlations of HRRPs, but not the target aspect information. To overcome the limitations in these feature learning methods, a new HRRP target recognition method that combines LSTM with HMM decision-making is proposed in this paper. The method consists of two branches: one is target recognition based on HMMs directly; the other is that the latent correlations of HRRPs are extracted by LSTM network and then use HMMs to do target recognition. Finally, the recognition result is obtained by making joint decisions between the two branches. The HRRPs are generated by the inversion of real radar images and the experimental results show that the proposed algorithm outperforms both the HMM and LSTM method.
Aircraft Recognition in Remote Sensing Images Based on Deep Learning

Jiaxi Lin,Xinde Li,Hong Pan

DOI: https://doi.org/10.1109/yac.2018.8406498

2018-01-01

Abstract:Object recognition is one of the fundamental issues in the field of computer vision. In traditional methods, invariant features are extracted from segmented targets for recognition. However, there is no common method for segmentation of aircraft targets so far due to the complex backgrounds, illuminations, noise and other practical factors. Therefore, in this paper, we propose a method for aircraft identification in remote sensing images based on HOG and deep learning features. We train two classifiers, one is the SVM classifier based on HOG feature, and the other is a classifier based on deep convolutional neural network VGGNet. First, we use the SVM classifier to identify the aircraft in the picture roughly, then we use the deep learning classifier to exclude misidentified targets. In this way, this coarse to fine framework can significantly improve the speed and accuracy of aircraft recognition in remote sensing images. At the same time, our method has a better generalization capability than the traditional methods. Experimental results demonstrate the robustness of our method.
An Efficient Target Detection and Recognition Method in Aerial Remote-sensing Images Based on Multiangle Regions-of-Interest

Guangcun Shan,Hongyu Wang,Wei Liang,Congcong Liu,Qizi Ma,Quan Quan

DOI: https://doi.org/10.48550/arXiv.1907.09320

2019-07-22

Computer Vision and Pattern Recognition

Abstract:Recently, deep learning technology have been extensively used in the field of image recognition. However, its main application is the recognition and detection of ordinary pictures and common scenes. It is challenging to effectively and expediently analyze remote-sensing images obtained by the image acquisition systems on unmanned aerial vehicles (UAVs), which includes the identification of the target and calculation of its position. Aerial remote sensing images have different shooting angles and methods compared with ordinary pictures or images, which makes remote-sensing images play an irreplaceable role in some areas. In this study, a new target detection and recognition method in remote-sensing images is proposed based on deep convolution neural network (CNN) for the provision of multilevel information of images in combination with a region proposal network used to generate multiangle regions-of-interest. The proposed method generated results that were much more accurate and precise than those obtained with traditional ways. This demonstrated that the model proposed herein displays tremendous applicability potential in remote-sensing image recognition.
Local Context Deformable Parts Model-Based Method for Remote Sensing Airport Aircraft Detection

Fukun Bi,Zhihua Yang,Mingyang Lei,Dan Huang,Liang Chen

DOI: https://doi.org/10.1117/1.jrs.13.046507

IF: 1.568

2019-01-01

Journal of Applied Remote Sensing

Abstract:As one of the most challenging tasks of optical remote sensing object detection, airport aircraft detection has attracted more attention from researchers. However, various scales and orientation of aircraft, and the complexity of airport regions leads to low detection performance. To address these problems, this article presents an effective aircraft detection method called the local context (LC) deformable parts model. First, aircraft candidate region fast extraction is developed to improve the detection efficiency. Then, the LC weight map is constructed by weighting the key points with the aircraft local characteristics. Afterward, gradient features are combined with the weight value to establish the LC histogram of oriented gradient feature pyramid. Finally, based on the established feature pyramid, aircraft candidate region is identified efficiently by orientation prediction of the suspected target. A series of comparison experiments on two datasets demonstrate that the proposed method has higher detection accuracy and efficiency than the existing detection methods. (C) 2019 Society of Photo-Optical Instrumentation Engineers (SPIE)
Automatic Target Recognition Method of Flight Vehicle Based on Template Matching

Huihui Li,Qili Chen,Junfang Fan

DOI: https://doi.org/10.1109/yac53711.2021.9486636

2021-01-01

Abstract:The key factor in obtaining flight vehicle guidance information is the automatic recognition of strapdown images. When using strap-down image guidance, it is more difficult to obtain guidance information due to the influence of environment and other factors on the information collection. Aiming at the fact that deep learning algorithms cannot be directly applied to the target recognition of strapdown images, a new target recognition framework is proposed. The framework extracts features from target template images and strapdown images, and uses the extracted feature maps as the input of the deep neural network, so that the improved target recognition framework can be directly applied to the strapdown environment. Experimental results show that the proposed framework can effectively recognize strapdown images, improve recognition accuracy, and achieve precise guidance.
Explicable Fine-Grained Aircraft Recognition Via Deep Part Parsing Prior Framework for High-Resolution Remote Sensing Imagery

Dingyuan Chen,Yanfei Zhong,Ailong Ma,Zhuo Zheng,Liangpei Zhang

DOI: https://doi.org/10.1109/TCYB.2023.3293033

Abstract:Aircraft recognition is crucial in both civil and military fields, and high-spatial resolution remote sensing has emerged as a practical approach. However, existing data-driven methods fail to locate discriminative regions for effective feature extraction due to limited training data, leading to poor recognition performance. To address this issue, we propose a knowledge-driven deep learning method called the explicable aircraft recognition framework based on a part parsing prior (APPEAR). APPEAR explicitly models the aircraft's rigid structure as a pixel-level part parsing prior, dividing it into five parts: 1) the nose; 2) left wing; 3) right wing; 4) fuselage; and 5) tail. This fine-grained prior provides reliable part locations to delineate aircraft architecture and imposes spatial constraints among the parts, effectively reducing the search space for model optimization and identifying subtle interclass differences. A knowledge-driven aircraft part attention (KAPA) module uses this prior to achieving a geometric-invariant representation for identifying discriminative features. Part features are generated by part indexing in a specific order and sequentially embedded into a compact space to obtain a fixed-length representation for each part, invariant to aircraft orientation and scale. The part attention module then takes the embedded part features, adaptively reweights their importance to identify discriminative parts, and aggregates them for recognition. The proposed APPEAR framework is evaluated on two aircraft recognition datasets and achieves superior performance. Moreover, experiments with few-shot learning methods demonstrate the robustness of our framework in different tasks. Ablation analysis illustrates that the fuselage and wings of the aircraft are the most effective parts for recognition.
LLE-based Classification Algorithm and Its Application to Passive mmW Target Recognition

Luo Lei,Li Yue-hua

DOI: https://doi.org/10.3724/SP.J.1146.2009.00877

2010-01-01

Abstract:A new one-class classification algorithm is put forward based on the idea of Locally Linear Embedding(LLE) and the low dimensional manifold of data distribution,for the issue of one-class classification of pattern recognition.The low dimensional manifold the short-time Fourier spectrum of passive millimeter wave(MMW) detector echo signal is found,which characteristics are explored as well.The algorithm is more efficient than present popular one-class classification algorithm when applied to passive MMW metal target recognition.Moreover,it is not sensitive to input parameters and is robust to data aliasing.
A Method of Target Recognition from Remote Sensing Images

Yili Fu,Kun Xing,Xianwei Han,Shuguo Wang

DOI: https://doi.org/10.1109/iros.2009.5354781

2009-01-01

Abstract:According to the characteristics of airfield and harbor from remote sensing images, a method of large target recognition based on the combination of target region and shape features is presented. First, edge detection and improved Hough transform are used to select line segments, the region including regular-array line segments in image is considered as region of interesting (ROI). ROI detection is the base for recognition. Target geometry shape is extracted from ROI using optimum threshold segmentation, which removes location effect and improves efficiency. As calculating shape principal orientations, all shapes are rotated to the same horizontally right to avoid rotation effect. The features extracted from shape implement multi-levels representation with moment features, normalized moment of inertia, length-width ratio and compact ration. Finally, feature vectors are normalized to measure similarity between target and template. Experiments show that target regions can be located accurately using ROI detection and it is effective for target recognition. Besides, the extracted features have good invariability with respect to rotation, translation and scaling, and they comprise local and overall consistency of the target, therefore, the recognition results meet expectations well.
LLE-based Classification Algorithm for MMW Radar Target Recognition

Lei Luo,Yuehua Li,Yinghong Luan

DOI: https://doi.org/10.1007/s11767-009-0093-4

2010-01-01

Journal of Electronics (China)

Abstract:In this paper, a new multiclass classification algorithm is proposed based on the idea of Locally Linear Embedding (LLE), to avoid the defect of traditional manifold learning algorithms, which can not deal with new sample points. The algorithm defines an error as a criterion by computing a sample’s reconstruction weight using LLE. Furthermore, the existence and characteristics of low dimensional manifold in range-profile time-frequency information are explored using manifold learning algorithm, aiming at the problem of target recognition about high range resolution MilliMeter-Wave (MMW) radar. The new algorithm is applied to radar target recognition. The experiment results show the algorithm is efficient. Compared with other classification algorithms, our method improves the recognition precision and the result is not sensitive to input parameters.
Remote sensing target recognition based on adaptive extreme learning machine

Nan Zhang,Shifei Ding,Xinzheng Xu

DOI: https://doi.org/10.13232/j.cnki.jnju.2014.04.012

2014-01-01

Abstract:Now,with the rapid progress of remote sensing technology,remote sensing target recognition has a very important practical significance and wide range of applications whether in civilian or military fields.However,due to high-resolution remote sensing images and other obj ective conditions,we can’t achieve real-time and accurate target recognition in the process of remote sensing image target recognition.Extreme learning machine(ELM)is a single-hidden layer feedforward neural network (SLFN )with at most N hidden nodes and with almost any nonlinear activation function can exactly learn N distinct observations.It should be noted that the input weights(linking the input layer to the first hidden layer)and hidden layer biases need to be adjusted in all these previous theoretical research works as well as in almost all practical learning algorithms of feedforward neural networks.ELM has the advantage of fast learning speed and is a one-time learning method,which has been widely used in small-sample learning problems.Aircraft and ships as typical artificial obj ects,having regular shape features,can be regarded as ideal geometric primitives,so aircraft and ship target can be seen as different category shapes.Firstly,to extract remote sensing images feature,and then ELM is applied to remote sensing target recognition,which is an effective way to solve the problem.However,ELM has the disadvantage of profuse hidden nodes (usually the same with number of learning samples )and huge model structure.Profuse hidden nodes inevitably lead to decrease the computation speed of the model.Therefore,the paper firstly puts forward a learning method based on ELM in which the number of hidden layer neurons is determined automatically - adaptive extreme learning machine (adaptive-ELM)which overcomes the disadvantage of profuse hidden nodes in ELM network,and then introduces remote sensing image feature extraction methods.The adaptive-ELM method firstly adaptively determines the number of hidden layer neurons through affinity propagation(AP)clustering sample set,and then make use of the ELM model to classify.AP is a clustering algorithm based on the concept of"message passing"between data points and the obj ective of AP is to find the optimal set of class representatives.Finally,simulation results are presented to demonstrate the accuracy and practicability of remote sensing target recognition based on adaptive-ELM.
Remote Sensing Aircraft Detection Method Based on Lightweight YOLOv4

Fei Cheng,Huanxin Zou,Xu Cao,Runlin Li,Shitian He,Juan Wei,Li Sun

DOI: https://doi.org/10.1117/12.2604561

2021-01-01

Abstract:Remote sensing aircraft target detection is an important task in the field of remote sensing images interpretation. The general target detection network often includes a large number of parameters, slow detection speed, and poor performance when directly applied to aircraft detection tasks. In order to solve the issues, a novel remote sensing aircraft target detection method based on lightweight YOLOv4 is proposed in this paper. Firstly, Lightweight YOLOv4 adopts the MobileNetV3 and depthwise separable convolution to greatly reduce the amount of model parameters. Then, to further enhance the feature extraction ability of the network and make the network more lightweight, this paper uses the feature enhancement module (FEM) and residual fusion module (RFM). Extensive experiments on the DOTA aircraft dataset demonstrate that the Lightweight YOLOv4 can significantly improve the detection accuracy and efficiency, as well as owns fewer model parameters.
Aircraft Recognition in Low-Resolution Sar Imagery Using Peak Feature Matching

Zhaodong Niu,Jiameng Pan,Chenglong Lin,Zengping Chen

DOI: https://doi.org/10.1117/12.2205838

2015-01-01

Abstract:Aircraft recognition is of great theoretical and practical significance in fields like remote sensing, navigation and traffic monitoring. It seems difficult to recognize aircraft in low-resolution SAR imagery because of difference between real image and simulated template induced by poor image quality and inherent simulation error. Aiming at this problem, an aircraft recognition method using peak feature matching is proposed. Firstly, the scattering centers of detected target are extracted in low-resolution SAR imagery using an adaptive threshold. Secondly, the extracted peak features are used to estimate the aircraft azimuth angle, which can be used to reduce the searching space in template database dramatically. Finally, a novel peak feature matching method using spatial distribution information of entire peak set is proposed to measure the similarity between detected target and simulated template. Experimental results demonstrate the good performance of the proposed method on a variety of low-resolution SAR imageries.
Radar Target Recognition Based on KLLE and a KNRD Classifier

Zhou Yun,Yu Xuelian,Liu Benyong,Wang Xuegang

2010-01-01

Abstract:This paper presents a radar target recognition method using kernel locally linear embedding (KLLE) and a kernel-based nonlinear representative and discriminative (KNRD) classifier. Locally linear embedding (LLE) is one of the representative manifold learning algorithms for dimensionality reduction. In this paper, LLE is extended by using kernel technique, which gives rises to the KLLE algorithm. A KNRD classifier is a combined version of a kernel-based nonlinear representor (KNR) and a kernel-based nonlinear discriminaor (KND), two classifiers recently proposed for optimal representation and discrimination, respectively. KLLE is firstly utilized to reduce data dimension and extract features from a high resolution range profile (HRRP). Then, a KNRD classifier is employed for classification. Experimental results on measured profiles from three aircrafts indicate the relatively good recognition performance of the presented method.

A METHOD OF AIRCRAFT TARGET RECOGNITION BASED ON LLE AND HMM

Automatic Aircraft Recognition using DSmT and HMM

Target Location Of Loitering Munitions Based On Image Matching

High Accuracy Ground Target Location Using Loitering Munitions Platforms

Automatic method for aircraft target recognition in remote sensing images based on improved active shape model

Aircraft Type Recognition Based on Target Track

Aircraft Recognition Based on Improved Iterative Threshold Selection and Skeleton Zernike Moment

Automatic Target Recognition Method Based on Sequential Images

A Novel HRRP Target Recognition Method Based on LSTM and HMM Decision-making.

Aircraft Recognition in Remote Sensing Images Based on Deep Learning

An Efficient Target Detection and Recognition Method in Aerial Remote-sensing Images Based on Multiangle Regions-of-Interest

Local Context Deformable Parts Model-Based Method for Remote Sensing Airport Aircraft Detection

Automatic Target Recognition Method of Flight Vehicle Based on Template Matching

Explicable Fine-Grained Aircraft Recognition Via Deep Part Parsing Prior Framework for High-Resolution Remote Sensing Imagery

LLE-based Classification Algorithm and Its Application to Passive mmW Target Recognition

A Method of Target Recognition from Remote Sensing Images

LLE-based Classification Algorithm for MMW Radar Target Recognition

Remote sensing target recognition based on adaptive extreme learning machine

Remote Sensing Aircraft Detection Method Based on Lightweight YOLOv4

Aircraft Recognition in Low-Resolution Sar Imagery Using Peak Feature Matching

Radar Target Recognition Based on KLLE and a KNRD Classifier