Search

Recent studies have shown that wireless mesh networks (WMNs) can be cheap, reliable, and efficient solutions for Internet of Things (IoTs) applications and connected devices. However, the increase in the size of the WMNs could lead to a degradation in performance. This makes the network vulnerable to high error rates over noisy and fading channels. This has increased the demand for more efficient channel coding schemes that can provide reliability, high data rate, high coding gain, energy efficiency, and minimum decoding complexity especially for low-cost and special purpose WMNs. In this research, an original multilevel coding with twofold concatenation scheme is designed based on high-rate space-time block codes (HRSTBCs) to cope with this increasing demand, particularly in MIMO-based WMNs. First, a multilevel coding scheme is implemented by correlating HRSTBCs with the uniquely constructed set-partitioning of the transmission matrix, based on the coding gain distance (CGD) criterion. This generates a new scheme, namely multilevel high-rate space-time block code (MHRSTBC), to provide a high transmission rate. Second, a twofold concatenation coding scheme is introduced by concatenating Reed-Solomon (RS) code, has been used for its energy efficiency and optimality in the mobile environment, with the inner code of the associated MHRSTBC. This has formed a new scheme that reaches the maximum coding gains, called the Reed-Solomon multilevel high-rate space-time block code (RS-MHRSTBC). The efficiency of the RS-MHRSTBC is verified by a computer simulation over a Rayleigh flat-fading and additive white Gaussian noise (AWGN) channel. The RS-MHRSTBC is compared with the classical schemes of Alamouti STBC and Multilevel STBC over uncoded QPSK, and the RS-MHRSTBC shows significant high coding gains reached of 47.79 dB, while Alamouti STBC and Multilevel STBC reached 38.12 dB and 44.30 dB at a BER of 10-6 respectively. RS-MHRSTBC also shows a reasonable decoding complexity while maintaining full transmission diversity at spectral efficiency of 2 bits/s/Hz.

Hyperspectral compressed sensing (HCS) based on spectral unmixing technique has shown great reconstruction performance. In particular, the linear mixed model (LMM) has been widely used in HCS reconstruction. However, due to the complexity of environmental conditions, instrumental configurations, and material nonlinear mixing effects, LMM cannot accurately represent the hyperspectral images, which limits the improvement of reconstruction quality. In this article, first, by introducing spectral variability, nonlinear mixing, and residuals, a multitype mixed model (MMM) is proposed to establish a more accurate hyperspectral image model. Then, a novel MMM-based HCS is proposed, which performs spectral compressed sampling at the sampling stage only, and at the reconstruction stage, by using spectral unmixing, an MMM-based HCS super-resolution reconstruction algorithm from spectral compressed sensing data is developed, and the alternating direction multiplier method is employed to estimate each component of the MMM, furthermore, reasonable prior knowledge of each component is introduced to improve the estimation accuracy. Experimental results on hyperspectral datasets demonstrate that the proposed model outperforms those state-of-the-art methods based on the LMM in terms of HCS reconstruction quality.

Due to the large data volume, the UAV image stitching and matching suffers from high computational cost. The traditional feature extraction algorithms—such as Scale-Invariant Feature Transform (SIFT), Speeded Up Robust Features (SURF), and Oriented FAST Rotated BRIEF (ORB)—require heavy computation to extract and describe features in high-resolution UAV images. To overcome this issue, You Only Look Once version 3 (YOLOv3) combined with the traditional feature point matching algorithms is utilized to extract descriptive features from the drone dataset of residential areas for roof detection. Unlike the traditional feature extraction algorithms, YOLOv3 performs the feature extraction solely on the proposed candidate regions instead of the entire image, thus the complexity of the image matching is reduced significantly. Then, all the extracted features are fed into Structural Similarity Index Measure (SSIM) to identify the corresponding roof region pair between consecutive image sequences. In addition, the candidate corresponding roof pair by our architecture serves as the coarse matching region pair and limits the search range of features matching to only the detected roof region. This further improves the feature matching consistency and reduces the chances of wrong feature matching. Analytical results show that the proposed method is 13× faster than the traditional image matching methods with comparable performance.

The dark channel prior (DCP)-based single image removal algorithm achieved excellent performance. However, due to the high complexity of the algorithm, it is difficult to satisfy the demands of real-time processing. In this article, we present a Graphics Processing Unit (GPU) accelerated parallel computing method for the real-time processing of high-definition video haze removal. First, based on the memory access pattern, we propose a simple but effective filter method called transposed filter combined with the fast local minimum filter algorithm and integral image algorithm. The proposed method successfully accelerates the parallel minimum filter algorithm and the parallel mean filter algorithm. Meanwhile, we adopt the inter-frame atmospheric light constraint to suppress the flicker noise in the video haze removal and simplify the estimation of atmospheric light. Experimental results show that our implementation can process the 1080p video sequence with 167 frames per second. Compared with single thread Central Processing Units (CPU) implementation, the speedup is up to 226× with asynchronous stream processing and qualified for the real-time high definition video haze removal.

Differentiation between benign and malignant breast cancer cases in X-ray images can be difficult due to their similar features. In recent studies, the transfer learning technique has been used to classify benign and malignant breast cancer by fine-tuning various pre-trained networks such as AlexNet, visual geometry group (VGG), GoogLeNet, and residual network (ResNet) on breast cancer datasets. However, these pre-trained networks have been trained on large benchmark datasets such as ImageNet, which do not contain labeled images related to breast cancers which lead to poor performance. In this research, we introduce a novel technique based on the concept of transfer learning, called double-shot transfer learning (DSTL). DSTL is used to improve the overall accuracy and performance of the pre-trained networks for breast cancer classification. DSTL updates the learnable parameters (weights and biases) of any pre-trained network by fine-tuning them on a large dataset that is similar to the target dataset. Then, the updated networks are fine-tuned with the target dataset. Moreover, the number of X-ray images is enlarged by a combination of augmentation methods including different variations of rotation, brightness, flipping, and contrast to reduce overfitting and produce robust results. The proposed approach has demonstrated a significant improvement in classification accuracy and performance of the pre-trained networks, making them more suitable for medical imaging.

Abstract: Real-time anomaly detection has received wide attention in remote sensing image processing because many moving targets must be detected on a timely basis. A widely-used anomaly detection algorithm is the Reed-Xiaoli (RX) algorithm that was proposed by Reed and Yu. The kernel RX algorithm proposed by Kwon and Nasrabadi is a nonlinear version of the RX algorithm and outperforms the RX algorithm in terms of detection accuracy. However, the kernel RX algorithm is computationally more expensive. This paper presents a novel real-time anomaly detection framework based on the kernel RX algorithm. In the kernel RX detector, the inverse covariance matrix and the estimated mean of the background data in the kernel space are non-causal and computationally inefficient. In this work, a local causal sliding array window is used to ensure the causality of the detection system. Using the matrix inversion lemma and the Woodbury matrix identity, both the inverse covariance matrix and estimated mean can be recursively derived without extensive repetitive calculations, and, therefore, the real-time kernel RX detector can be implemented and processed pixel-by-pixel in real time. To substantiate its effectiveness and utility in real-time anomaly detection, real hyperspectral data sets are utilized for experiments.