Search

Endmember extraction is one of the necessary steps in hyperspectral data investigation. The eventual objective of hyperspectral data processing and analysis is to improve the accuracy of target identification, and the precise identification of endmembers is a challenging task. The several of the algorithms proposed for endmember extraction require knowledge on the targets for appropriate initialization. The endmember extraction algorithms (EEAs) rely on endmember initialization algorithms (EIAs) and in turn many EIAs also require the knowledge on targets of interest to begin the search process effectively. A comprehensive study on targets of interest to begin the search process is due expected. Therefore, in this paper, the concept of extreme or boundary pixels is explored to identify the targets of interest and the darkest pixel (a pixel with minimum length) is identified and proposed as a new target of interest to begin the endmember search process. The utility/influence of the proposed target of interest, i.e. the darkest pixel, is studied with that's of the brightest pixel (a pixel with maximum length) which has been in use as a popular target of interest for EIAs so far. An automatic target generation process (ATGP) and a similarity measures based endmember initialization algorithm (SMEIA) are adapted to test the proposed initialization strategy. The experimental results have revealed the usefulness of the darkest pixel as an additional target of interest in addition to the brightest pixel in searching for endmembers by EIAs. The strategy of choosing the darkest pixel enhanced the initial target knowledge and also improved the performance of the EIAs considerably.

Object tracking is being utilized in the field of computer vision over decades for video surveillance, human–computer interaction and robotic applications. Even though the state-of-the-art tracking...

Object tracking is a fundamental task in video surveillance, human-computer interaction and activity analysis. One of the common challenges in visual object tracking is illumination variation. A large number of methods for tracking have been proposed over the recent years, and median flow tracker is one of them which can handle various challenges. Median flow tracker is designed to track an object using Lucas-Kanade optical flow method which is sensitive to illumination variation, hence fails when sudden illumination changes occur between the frames. In this paper, we propose an enhanced median flow tracker to achieve an illumination invariance to abruptly varying lighting conditions. In this approach, illumination variation is compensated by modifying the Discrete Cosine Transform (DCT) coefficients of an image in the logarithmic domain. The illumination variations are mainly reflected in the low-frequency coefficients of an image. Therefore, a fixed number of DCT coefficients are ignored. Moreover, the Discrete Cosine (DC) coefficient is maintained almost constant all through the video based on entropy difference to minimize the sudden variations of lighting impacts. In addition, each video frame is enhanced by employing pixel transformation technique that improves the contrast of dull images based on probability distribution of pixels. The proposed scheme can effectively handle the gradual and abrupt changes in the illumination of the object. The experiments are conducted on fast-changing illumination videos, and results show that the proposed method improves median flow tracker with outperforming accuracy compared to the state-of-the-art trackers

Hyperspectral imaging is one of the most information-rich sources of remote sensing data that exists. It can capture the entire, continuous spectrum of color and light. Feature extraction techniques that are selected for identifying diagnostic features influence classification accuracy. Spectral matching algorithms like similarity measures are developed to compare spectral features of materials with their reference spec-tral signatures in identifying different earth surfaces. Similarity measures are used as simple feature extraction techniques in target identification using hyperspectral data. Though there are several similarity measures, selecting a robust similarity measure requires further investigation. Influence of similarity measures are not studied much in distinguishing spectrally distinct signatures. In this article, we propose to study the performance of similarity measures in (i) discriminating endmember signatures (ii) mixed pixel identification (iii) clustering signatures of different classes and (iv) endmember extraction. Experimental results show an effective and a robust performance of proposed phase correlation similarity measure among all other similarity measures compared for all the problems under investigation.

The classical theory of sampling offers an accurate reconstruction of band-limited signals, however, not for the non band-limited. Over the previous decade, Finite Rate of Innovation (FRI) framework has emerged that has overcome this limitation. The FRI framework considers the sampling and reconstruction of particular classes of signals which are non-bandlimited and are fully specified by a finite number of degrees of freedom per unit interval. Traditional FRI algorithms, such as matrix pencil and Cadzow algorithm followed by annihilating filter methods use the singular value decomposition (SVD) for signal reconstruction. Eventhough, these algorithms achieve optimal results, however, in the low signal-to-noise ratio (SNR) they become unstable due to the larger magnitudes of the noise singular values than signal singular values. In this paper, to overcome this problem, a residual convolutional neural network approach, which estimates the signal parameters, is proposed. This network learns from the training data and gives the signal parameters from the discrete sample values. Simulation results show significant improvements on the smaller SNR values over the traditional FRI algorithms.

In multispectral and hyperspectral remote sensing, classification of pixels is obtained by means of spectral similarity of known field or library spectra to unknown image spectra. Endmember extraction is the most decisive task in hyperspectral image analysis. Endmember initialization algorithms (EIAs) play a key role and support endmember extraction algorithms (EEAs) in extracting near optimal set of endmembers. Though there are few endmember initialization techniques available, similarity measures are not explored in detail in target generation. Hence, in this paper, it is proposed to explore similarity measures in identifying spectrally distinct signatures to use them as initial endmembers. Individual similarity measures are combined to form hybrid similarity measures to confirm their effectiveness in generating spectrally distinct targets. Initial set of endmembers extracted by proposed algorithm are used for initializing classical EEA, the NFINDR, which is sensitive to endmember initialization, and their performance in final endmembers selection is verified. Experimental results on two hyperspectral data sets show the superior performance of the similarity based endmember initialization algorithm (SMEIA).

This paper proposes a dual-band Antipodal Vivaldi antenna for wireless neural monitoring system. Most of the work in the literature uses active implant systems operating in single frequency band of operation. The proposed antenna utilizes a passive Antipodal Vivaldi shaped structure that achieves a dual band of operation. The proposed wireless neural monitoring system employs a microwave back-scattering method and eliminates the need for additional power sources. The Antipodal shaped antenna with optimized design provides two different bands with transmission at 4.4 GHz and reception at 2.2 GHz. Measured and simulated return loss results of the antenna shows a good agreement with each other. In addition, proposed antenna achieves the gains of 3.05 dB and 1.39 dB at dual-band while radiation pattern of the antenna is found more directive by achieving a 16 dB difference between co-polarization and cross-polarization.

Antennas in radar systems are protected on the outside by dielectric materials called radomes. Radomes are prone to boresight error and loss of electromagnetic energy. Metamaterials can be used in enhancing the performance of radome by eliminating boresight error and alleviating loss of energy. This paper proposes a compact single layer radome made up of an S-shaped zero-index metamaterial that operates at X-band frequency i.e., 10 GHz. A radome is designed by the periodic repetition of 10 x 7 S-ring unit cells. Radome is realized on Rogers RT/Duriod 5870 substrate (having permittivity ɛr = 2.33) with dimensions of 35 mm x 35mm x 0.5mm. The proposed design provides an overall gain of 9.8dB (18% more than antenna), eliminates the boresight error and has a wider band of operation from 9 to 11 GHz.

Biomedical image segmentation is one of the main and fast growing field in medical image processing domain. Deep neural networks is one of the popular field used for image segmentation. Convolutional neural networks(CNNs) in deep neural networks have shown good performance for biomedical image segmentation. However, a strong notion exists that large number of annotated images are required for training of CNNs. Therefore, in this paper we have come up with a modified U-Net architecture for limited number of annotated data with an intersection over union score of 92.54%. The architecture uses rectified-adam optimizer(advanced version of adam) for minimizing the loss function which helps us to come close to global optima. We have also compared the performance of various optimizers on the proposed network.

The intima media thickness (IMT) of common carotid artery is a reliable measure of cardiovascular diseases. The quantification of IMT is the biomarker for clinical diagnosis of the risk of stroke. For robust measurement of IMT, the ultrasound carotid images must be free of speckle noise. To reduce the effect of speckle noise in the carotid ultrasound image, we propose to use Bayesian least square estimation filter. In addition, the enhancement step based on total variation- $$L_{1}$$ (TV- $$L_{1}$$ ) norm is performed to improve the robustness. Further more, we present a fully automated region of interest and segmentation of intima media complex based on support vector machine. The quantitative evaluation is carried out on 49 carotid ultrasound images. The proposed algorithm is compared with gradient-based methods like model based, dynamic programming, snake algorithm, and classifier-based segmentation using a neural network algorithm. The performance of the experimental result shows that the proposed method is robust in quantifying the IMT in carotid ultrasound images.

This paper proposes a robust framework for quality restoration of remotely sensed aerial images. Proposed framework works in three steps: (1) Efficient color balancing and saturation adjustment, (2) Efficient color restoration, (3) Modified contrast enhancement using particle swarm optimization (PSO). In order to show the robustness, step-wise results of proposed framework is illustrated. Several aerial images from two publically available datasets are tested to support the robustness of the proposed framework over existing image quality restoration methods. The experimental results of proposed framework and other existing quality restoration methods are compared in terms of NIQMC, BIQME, MICHELSON, DE, EME and PIXDIST along with visual experimental results. Based on experimental results conducted on several aerial images suggest that the proposed framework is outperform over existing quality restoration methods.

The correlation filters (CF) have been extensively used in object tracking due to its robustness and attractive computational speed. However, the CF are more sensitive to object deformation because...

Edge detection is a primary image processing technique used for object detection, data extraction, and image segmentation. Recently, edge-based segmentation using structured classifiers has been receiving increasing attention. The intima media thickness (IMT) of the common carotid artery is mainly used as a primitive indicator for the development of cardiovascular disease. For efficient measurement of the IMT, we propose a fast edge-detection technique based on a structured random forest classifier. The accuracy of IMT measurement is degraded owing to the speckle noise found in carotid ultrasound images. To address this issue, we propose the use of a state-of-the-art denoising method to reduce the speckle noise, followed by an enhancement technique to increase the contrast. Furthermore, we present a novel approach for an automatic region of interest extraction in which a pre-trained structured random forest classifier algorithm is applied for quantifying the IMT. The proposed method exhibits IMTmean ± standard deviation of 0.66mm ± 0.14, which is closer to the ground truth value 0.67mm ± 0.15 as compared to the state-of-the-art techniques.

Cardiovascular diseases are the third leading cause of death worldwide. The primitive indication of the possible onset of a cardiovascular disease is atherosclerosis, which is the accumulation of plaque on the arterial wall. The intima-media thickness (IMT) of the common carotid artery is an early marker of the development of cardiovascular disease. The computation of the IMT and the delineation of the carotid plaque are significant predictors for the clinical diagnosis of the risk of stroke. For a robust diagnosis, carotid ultrasound images must be free from speckle noise. To address this problem, we use state-of-the-art despeckling and enhancement methods in this work. Many edge-based methods for IMT estimation have been proposed to overcome the limitations of manual segmentation. In this paper, we present a fully automated region-of-interest (ROI) extraction and a threshold-based segmentation of the intima media complex (IMC) using a wind driven optimization (WDO) technique. A quantitative evaluation is carried out on 90 carotid ultrasound images of two different datasets. The obtained results are compared with those of state-of-the-art techniques such as a model-based approach, a dynamic programming method, and a snake segmentation method. The experimental analysis shows that the proposed method is robust in measuring the IMT in carotid ultrasound images.

The process of designing an efficient tracker for thermal infrared imagery is one of the most challenging tasks in computer vision. Although a lot of advancement has been achieved in RGB videos over the decades, textureless and colorless properties of objects in thermal imagery pose hard constraints in the design of an efficient tracker. Tracking of an object using a single feature or a technique often fails to achieve greater accuracy. Here, we propose an effective method to track an object in infrared imagery based on a combination of discriminative and generative approaches. The discriminative technique makes use of two complementary methods such as kernelized correlation filter with spatial feature and AdaBoost classifier with pixel intesity features to operate in parallel. After obtaining optimized locations through discriminative approaches, the generative technique is applied to determine the best target location using a linear search method. Unlike the baseline algorithms, the proposed method estimates the scale of the target by Lucas-Kanade homography estimation. To evaluate the proposed method, extensive experiments are conducted on 17 challenging infrared image sequences obtained from LTIR dataset and a significant improvement of mean distance precision and mean overlap precision is accomplished as compared with the existing trackers. Further, a quantitative and qualitative assessment of the proposed approach with the state-of-the-art trackers is illustrated to clearly demonstrate an overall increase in performance.

Lane detection algorithms have been the key enablers for a fully-assistive and autonomous navigation systems. In this paper, a novel and pragmatic approach for lane detection is proposed using a convolutional neural network (CNN) model based on SegNet encoder-decoder architecture. The encoder block renders low-resolution feature maps of the input and the decoder block provides pixel-wise classification from the feature maps. The proposed model has been trained over 2000 image data-set and tested against their corresponding ground-truth provided in the data-set for evaluation. To enable real-time navigation, we extend our model's predictions interfacing it with the existing Google APIs evaluating the metrics of the model tuning the hyper-parameters. The novelty of this approach lies in the integration of existing segNet architecture with google APIs. This interface makes it handy for assistive robotic systems. The observed results show that the proposed method is robust under challenging occlusion conditions due to pre-processing involved and gives superior performance when compared to the existing methods., Preprint: To be published in the proceedings of IEEE TENCON 2019