Discriminative Feature Learning With Distance Constrained Stacked Sparse Autoencoder for Hyperspectral Target Detection

Shi, Yanzi; Lei, Jie; Yin, Yunfei; Cao, Kailang; Li, Yunsong; Chang, Chein‐I

doi:10.1109/lgrs.2019.2901019

Cited by 39 publications

(14 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…and DCSSAED [11]. The three real hyperspectral datasets adopted are HYDICE, AVIRIS and AVIRIS2 [5,8].…”

Section: Experimental Results and Analysismentioning

confidence: 99%

“…Earlier works adopt the spectrum as the representation of pixels directly and detect the target pixels by simple signal processing or sparse coding algorithms [9,10]. Recently, inspired by the great success of deep learning in computer vision tasks, deep neural networks have been employed to learn more discriminative representations [11]. Although better performance is achieved, these approaches also suffer from the problems of lacking annotated data as well as extremely sample imbalance.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Self-supervised spectral matching network for hyperspectral target detection

Yao¹,

Yuan²,

Jiang³

2021

Preprint

View full text Add to dashboard Cite

Hyperspectral target detection is a pixel-level recognition problem. Given a few target samples, it aims to identify the specific target pixels such as airplane, vehicle, ship, from the entire hyperspectral image. In general, the background pixels take the majority of the image and complexly distributed. As a result, the datasets are weak annotated and extremely imbalanced. To address these problems, a spectral mixing based self-supervised paradigm is designed for hyperspectral data to obtain an effective feature representation. The model adopts a spectral similarity based matching network framework. In order to learn more discriminative features, a pair-based loss is adopted to minimize the distance between target pixels while maximizing the distances between target and background. Furthermore, through a background separated step, the complex unlabeled spectra are downsampled into different sub-categories. The experimental results on three real hyperspectral datasets demonstrate that the proposed framework achieves better results compared with the existing detectors.

show abstract

“…and DCSSAED [11]. The three real hyperspectral datasets adopted are HYDICE, AVIRIS and AVIRIS2 [5,8].…”

Section: Experimental Results and Analysismentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Self-supervised spectral matching network for hyperspectral target detection

Yao¹,

Yuan²,

Jiang³

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…One approach is to extract the in-scene target signatures by endmember extraction. This approach has been proved beneficial to detection performance [45]. However, endmember extraction algorithms tend not to work well when there are few targets in the image.…”

Section: Discussionmentioning

confidence: 99%

“…The size of the IWR is fixed as above mentioned 15, and the range of the OWR size is set to [21,23,25,27,29,31,33,35,37]. The range of θ is set as [0, 0.1, 0.5, 1, 2, 3, 5, 10, 15] degree, and the range of K is set to [5,10,15,20,25,30,35,40,45,50]. Detection performance is evaluated by the AUC value.…”

Section: Parameter Analysismentioning

confidence: 99%

Joint Hapke Model and Spatial Adaptive Sparse Representation with Iterative Background Purification for Martian Serpentine Detection

Xia

Mustard

et al. 2021

Remote Sensing

View full text Add to dashboard Cite

Visible and infrared imaging spectroscopy have greatly revolutionized our understanding of the diversity of minerals on Mars. Characterizing the mineral distribution on Mars is essential for understanding its geologic evolution and past habitability. The traditional handcrafted spectral index could be ambiguous as it may denote broad mineralogical classes, making this method unsuitable for definitive mineral investigation. In this work, the target detection technique is introduced for specific mineral mapping. We have developed a new subpixel mineral detection method by joining the Hapke model and spatially adaptive sparse representation method. Additionally, an iterative background dictionary purification strategy is proposed to obtain robust detection results. Laboratory hyperspectral image containing Mars Global Simulant and serpentine mixtures was used to evaluate and tailor the proposed method. Compared with the conventional target detection algorithms, including constrained energy minimization, matched filter, hierarchical constrained energy minimization, sparse representation for target detection, and spatially adaptive sparse representation method, the proposed algorithm has a significant improvement in accuracy about 30.14%, 29.67%, 29.41%, 9.13%, and 8.17%, respectively. Our algorithm can detect subpixel serpentine with an abundance as low as 2.5% in laboratory data. Then the proposed algorithm was applied to two well-studied Compact Reconnaissance Imaging Spectrometer for Mars images, which contain serpentine outcrops. Our results are not only consistent with the spatial distribution of Fe/Mg phyllosilicates derived by spectral indexes, but also denote what the specific mineral is. Experimental results show that the proposed algorithm enables the search for subpixel, low-abundance, and scientifically valuable mineral deposits.

show abstract

“…Benefited from abundant spectral information, HSI has become instrumental in many application scenarios, such as medical diagnosis and treatment [1]- [3], agricultural production [4]- [6] and identification of the terrain landform [7]- [9]. Among these applications, target detection is considered as a fundamental task and has received a surge of interest [10], [11]. Essentially, target detection can be regarded as a problem of classification and localization [12], [13], which has been widely used spanning from civilian to military.…”

Section: Introductionmentioning

confidence: 99%

Background Learning Based on Target Suppression Constraint for Hyperspectral Target Detection

Xie

Zhang

et al. 2020

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

Self Cite

View full text Add to dashboard Cite

Hyperspectral target detection is critical in both military and civilian applications. However, it is a challenging task due to the complexity of background and the limited samples of target in hyperspectral images (HSIs). In this paper, we propose a novel background learning model, called background learning based on target suppression constraint (BLTSC) to characterize high-dimensional spectral vectors. Considering insufficient target samples, the model is trained only on the background spectral samples to accurately learn the background distribution. Then the discrepancy between the reconstructed and original HSIs are examined to spot the targets. To obtain a background training dataset, coarse detection is carried out. However, it is quite difficult to retrieve pure background data. Thus a target suppression constraint is imposed to reduce the impact of suspected target samples on background reconstruction. Experiments on six real HSIs demonstrate that the proposed framework significantly outperforms the current state-of-the-art detection methods and yields higher detection accuracy and lower false alarm rate. Index Terms-Hyperspectral image (HSI), target detection, background learning, target suppression constraint.

show abstract

Discriminative Feature Learning With Distance Constrained Stacked Sparse Autoencoder for Hyperspectral Target Detection

Cited by 39 publications

References 16 publications

Self-supervised spectral matching network for hyperspectral target detection

Self-supervised spectral matching network for hyperspectral target detection

Joint Hapke Model and Spatial Adaptive Sparse Representation with Iterative Background Purification for Martian Serpentine Detection

Background Learning Based on Target Suppression Constraint for Hyperspectral Target Detection

Contact Info

Product

Resources

About