Multistage Fusion With Dissimilarity Regularization for SAR/IR Target Recognition

Cho, Young-Rae; Shin, Seung‐Jun; Yim, Sung-Hyuk; Kong, Kyeongbo; Cho, Hyun-Woong

doi:10.1109/access.2018.2885736

Cited by 8 publications

(6 citation statements)

References 45 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Late fusion methods have been widely used in many fields because these methods can choose the state-of-the-art subnetworks [20], [21]. Of course, decision fusion also has limitations; using only the final vector from classifiers independently will cause loss of spatial information contained in the feature maps, which results in inadequate information use [22]. In summary, all fusion strategies have both advantages and limitations, so the choice of fusion network must be made in the context of the data and task at hand.…”

Section: Introductionmentioning

confidence: 99%

Automatic Diagnosis of Familial Exudative Vitreoretinopathy Using a Fusion Neural Network for Wide-Angle Retinal Images

Mao

Liu

et al. 2020

IEEE Access

View full text Add to dashboard Cite

Familial exudative vitreoretinopathy (FEVR) is a hereditary disorder that can damage the retina. This retinal damage can lead to vision loss and even blindness in the late stages. Thus, early diagnosis and prevention of the disease's progression are critical. The purpose of this study was to develop an automated diagnosis system for FEVR based on combining deep learning and domain knowledge. A transfer learning scheme was designed to train a deep convolutional neural network (DCNN) to provide segmentation of the retinal vessels. Based on this vessel segmentation and prior clinical knowledge, the vascular characteristics, including the retinal avascular area, vessel angle, fractal dimension, branching and density of blood vessels, were automatically evaluated. Finally, the diagnosis of FEVR was achieved by a feature fusion neural network. Our method was evaluated on 300 images with 168 healthy and 132 FEVR images. By combining deep features and handcrafted features (extracted vascular characteristics), the proposed method achieved an average F1-score of 0.95, with excellent accuracy (94.34%) and sensitivity (91.43%); the quadratic weighted κ was 0.88 for the diagnosis of FEVR. We demonstrated the effectiveness and robustness of the proposed method using five-fold cross-validation. The proposed automatic diagnosis system can assist doctors for better judgment and make sense of early diagnosis and prevention of the disease's progression. INDEX TERMS FEVR, wide-angle retinal images, deep learning, vascular characteristics, fusion.

show abstract

Section: Introductionmentioning

confidence: 99%

Automatic Diagnosis of Familial Exudative Vitreoretinopathy Using a Fusion Neural Network for Wide-Angle Retinal Images

Mao

Liu

et al. 2020

IEEE Access

View full text Add to dashboard Cite

show abstract

“…High-quality SAR images can better reflects the characteristics of the recognition target. In order to facilitate subsequent recognition operations and reduce the computational burden of the recognition algorithm, the use of machine vision imaging technology to collect SAR images enables the recognition algorithm to have relatively high recognition efficiency and recognition accuracy [16].…”

Section: A Sar Image Acquisition Based On Machine Visionmentioning

confidence: 99%

Recognition Method of Dim and Small Targets in SAR Images based on Machine Vision

Dong¹

2022

IJACSA

View full text Add to dashboard Cite

Aiming at the problems of long recognition time and low recognition accuracy of traditional SAR image dim target recognition methods, a method of SAR image dim target recognition based on machine vision was proposed. SAR images are collected and preprocessed by machine vision, and the image information is processed by PCA dimension reduction considering the linear characteristics of the data to extract image features. Then, the SAR image target feature key frame frequency band is divided by the segmentation results, and the recognition model is established based on the image trajectory tracking and target analysis. The proposed algorithm is applied and analyzed. The simulation results show that the proposed algorithm has good recognition rate, average recognition rate and false detection rate are 99% and 0.9%, and can effectively ensure the data processing performance.

show abstract

“…This framework achieved 99.33% and 99.86% for the three and ten-class problems on MSTAR, respectively. A SAR and infrared (IR) sensors based multistage fusion stream strategy with dissimilarity regularization using CNN architecture was developed in [363] to improve the performance of SAR target recognition. In order to make full use of phase information of PolSAR images and extract more robust discriminative features with multidirection, multiscale, and multiresolution properties, a complex Contourlet CNN was proposed in [376].…”

Section: A Sar Images Processingmentioning

confidence: 99%

A Comprehensive Survey of Machine Learning Applied to Radar Signal Processing

Lang,

Fu,

Martorella

et al. 2020

Preprint

View full text Add to dashboard Cite

Modern radar systems have high requirements in terms of accuracy, robustness and real-time capability when operating on increasingly complex electromagnetic environments. Traditional radar signal processing (RSP) methods have shown some limitations when meeting such requirements, particularly in matters of target classification. With the rapid development of machine learning (ML), especially deep learning, radar researchers have started integrating these new methods when solving RSP-related problems. This paper aims at helping researchers and practitioners to better understand the application of ML techniques to RSP-related problems by providing a comprehensive, structured and reasoned literature overview of ML-based RSP techniques. This work is amply introduced by providing general elements of ML-based RSP and by stating the motivations behind them. The main applications of ML-based RSP are then analysed and structured based on the application field. This paper then concludes with a series of open questions and proposed research directions, in order to indicate current gaps and potential future solutions and trends.

show abstract

Multistage Fusion With Dissimilarity Regularization for SAR/IR Target Recognition

Cited by 8 publications

References 45 publications

Automatic Diagnosis of Familial Exudative Vitreoretinopathy Using a Fusion Neural Network for Wide-Angle Retinal Images

Automatic Diagnosis of Familial Exudative Vitreoretinopathy Using a Fusion Neural Network for Wide-Angle Retinal Images

Recognition Method of Dim and Small Targets in SAR Images based on Machine Vision

A Comprehensive Survey of Machine Learning Applied to Radar Signal Processing

Contact Info

Product

Resources

About