skip to main content

Title: Efficient hyperspectral target detection using class-associative spectral fringe- adjusted JTC with dimensionality reduction techniques
Recent studies have shown that fringe-adjusted joint transform correlation (FJTC) can be effectively applied for single class and even multiclass object detection in hyperspectral imagery (HSI). However, directly utilizing FJTC based techniques for HSI processing may not be efficient due to the fact that HSI may contain a large volume of data redundancy. Therefore, incorporating dimensionality reduction (DR) methods prior to the object detection procedure is suggested. In this paper, we combine several DRs individually with class-associative spectral FJTC (CSFJTC), and then compare their performance on single class and multiclass object detection tasks using a real-world hyperspectral data set. Test results show that the CSFJTC with denoising autoencoder provides superior performance compared to the alternate methods for detecting few dissimilar patterns in the scene.
Award ID(s):
Publication Date:
Journal Name:
Asian Journal of Physics
Page Range or eLocation-ID:
Sponsoring Org:
National Science Foundation
More Like this
  1. null (Ed.)
    We propose a method for the unsupervised clustering of hyperspectral images based on spatially regularized spectral clustering with ultrametric path distances. The proposed method efficiently combines data density and spectral-spatial geometry to distinguish between material classes in the data, without the need for training labels. The proposed method is efficient, with quasilinear scaling in the number of data points, and enjoys robust theoretical performance guarantees. Extensive experiments on synthetic and real HSI data demonstrate its strong performance compared to benchmark and state-of-the-art methods. Indeed, the proposed method not only achieves excellent labeling accuracy, but also efficiently estimates the number of clusters. Thus, unlike almost all existing hyperspectral clustering methods, the proposed algorithm is essentially parameter-free.
  2. Hyperspectral imaging (HSI) technology has been applied in a range of fields for target detection and mixture analysis. While its original applications were in remote sensing, modern uses include agriculture, historical document authentications and medicine. HSI has shown great utility in fluorescence microscopy; however, acquisition speeds have been slow due to light losses associated with spectral filtering. We are currently developing a rapid hyperspectral imaging platform for 5-dimensional imaging (RHIP-5D), a confocal imaging system that will allow users to obtain simultaneous measurements of many fluorescent labels. We have previously reported on optical modeling performance of the system. This previous model investigated geometrical capability of designing a multifaceted mirror imaging system as an initial approach to sample light at many wavelengths. The design utilized light-emitting diodes (LEDs) and a multifaceted mirror array to combine light sources into a liquid light guide (LLG). The computational model was constructed using Monte Carlo optical ray software (TracePro, Lambda Research Corp.). Recent results presented here show transmission has increased up to 9% through parametric optimization of each component. Future work will involve system validation using a prototype engineered based on our optimized model. System requirements will be evaluated to determine if potential design changes aremore »needed to improve the system. We will report on spectral resolution to demonstrate feasibility of the RHIP-5D as a promising solution for overcoming current HSI acquisition speed and sensitivity limitations.« less
  3. Hyperspectral imagery (HSI) has emerged as a highly successful sensing modality for a variety of applications ranging from urban mapping to environmental monitoring and precision agriculture. Despite the efforts by the scientific community, developing reliable algorithms of HSI classification remains a challenging problem especially for high-resolution HSI data where there is often larger intraclass variability combined with scarcity of ground truth data and class imbalance. In recent years, deep neural networks have emerged as a promising strategy for problems of HSI classification where they have shown a remarkable potential for learning joint spectral-spatial features efficiently via backpropagation. In this paper, we propose a deep learning strategy for HSI classification that combines different convolutional neural networks especially designed to efficiently learn joint spatial-spectral features over multiple scales. Our method achieves an overall classification accuracy of 66.73% on the 2018 IEEE GRSS hyperspectral dataset – a high-resolution dataset that includes 20 urban land-cover and land-use classes
  4. Agaian, Sos S. ; Jassim, Sabah A. ; DelMarco, Stephen P. ; Asari, Vijayan K. (Ed.)
    Neural networks have emerged to be the most appropriate method for tackling the classification problem for hyperspectral images (HIS). Convolutional neural networks (CNNs), being the current state-of-art for various classification tasks, have some limitations in the context of HSI. These CNN models are very susceptible to overfitting because of 1) lack of availability of training samples, 2) large number of parameters to fine-tune. Furthermore, the learning rates used by CNN must be small to avoid vanishing gradients, and thus the gradient descent takes small steps to converge and slows down the model runtime. To overcome these drawbacks, a novel quaternion based hyperspectral image classification network (QHIC Net) is proposed in this paper. The QHIC Net can model both the local dependencies between the spectral channels of a single-pixel and the global structural relationship describing the edges or shapes formed by a group of pixels, making it suitable for HSI datasets that are small and diverse. Experimental results on three HSI datasets demonstrate that the QHIC Net performs on par with the traditional CNN based methods for HSI Classification with a far fewer number of parameters. Keywords: Classification, deep learning, hyperspectral imaging, spectral-spatial feature learning
  5. Hyperspectral imaging (HSI) is a spectroscopic technique which captures images at a high contrast over a wide range of wavelengths to show pixel specific composition. Traditional uses of HSI include: satellite imagery, food distribution quality control and digital archaeological reconstruction. Our lab has focused on developing applications of HSI fluorescence imaging systems to study molecule-specific detection for rapid cell signaling events or real-time endoscopic screening. Previously, we have developed a prototype spectral light source, using our modified imaging technique, excitationscanning hyperspectral imaging (HIFEX), coupled to a commercial colonoscope for feasibility testing. The 16 wavelength LED array was combined, using a multi-branched solid light guide, to couple to the scope’s optical input. The prototype acquired a spectral scan at near video-rate speeds (~8 fps). The prototype could operate at very rapid wavelength switch speeds, limited to the on/off rates of the LEDs (~10 μs), but imaging speed was limited due to optical transmission losses (~98%) through the solid light guide. Here we present a continuation of our previous work in performing an in-depth analysis of the solid light guide to optimize the optical intensity throughput. The parameters evaluated include: LED intensity input, geometry (branch curvature and combination) and light propagation usingmore »outer claddings. Simulations were conducted using a Monte Carlo ray tracing software (TracePro). Results show that transmission within the branched light guide may be optimized through LED focusing lenses, bend radii and smooth tangential branch merges. Future work will test a new fabricated light guide from the optimized model framework.« less