Search

Representation and classification of multi-dimensional data are current key research areas. The representation of data in two classes is more feasible than multi-class representations because of the inherent quadratic complexity in existing techniques. Erroneous assignment of class labels affects separation boundary and training time complexity. In this paper, multi-dimensional data is handled using linear discriminant analysis (LDA) and threefold support vector machine (SVM) techniques to reduce the complexity and minimize false labeling. A facial expression application is proposed in which six natural expressions are used as multi-class data. Face image is divided into seven triangles on the basis of two focal points. A combined local and global feature descriptor is generated. Discrete Fourier transform is applied and processed with LDA to obtain discriminant features and accurately map an input feature space to an output space. To evaluate the system performance, Japanese Female Facial Expression, FER-2013 and Cohn–Kanade DFAT datasets are used. The obtained results show that multi-class data hyper plane using LDA and threefold SVM approach is effective and simple for quadratic data analysis

Hazy image observation is a serious phenomenon occurred by random scattering of atmospheric elements. This reduces the scene prominence in terms of color contrast and makes the vision system poor. To overcome this issue, dehazing of digital images is dealt with much attention in the field of computer vision. Dense scattering in hazy image results in loss of original spectral information of the object. It is required to remove the unwanted artifacts of weather during image acquisition process of the given object to improve the visual insight of an acquired image. It is found that most of the existing methods fail to retrieve the sufficient information from hazy images. To address such problem, a robust single image dehazing method is introduced to recover the haze-free image. This is performed by calculating the atmospheric light of the given hazy images by decomposing and retaining the high frequency sub bands using wavelet domain. The dense haze elimination is performed on approximated low frequency sub band of the given hazy image. On reconstruction process, well-refined dehaze image is obtained. Moreover, transmission maps are also produced as a byproduct of this method.

Lung cancer is a fatal type of cancer and it causes of severe deaths of approximately 422 people every day, worldwide. However, an early diagnosis is an expedient requirement for increasing the chances of human survival. In this regard the existing techniques of tumor detection, CT scans are mostly utilized to recognize the infected regions, nevertheless, the major challenges of CT images are low visibility of tumor regions, negative rates, to name but a few. In this work, we propose a novel design of contrast stretching based classical features fusion process for localizing the of lungs cancer classification. The proposed method encompasses three significant steps: firstly, contrast of original CT images is improved by gamma correction max intensity weights approach. Secondly, multiple texture, point, and geometric features are extracted from contrast images, which are later subjected to a serial canonical correlation-based fusion. Thirdly, zero values and negative features are replaced by an entropy-based approach, followed by weighted NCA for selection. Most discriminate features are fed into ensemble classifier for final classification. The validation of the proposed method is conducted on publicly available dataset: Lungs Data Science Bowl 2017 to achieving maximum accuracy of 99.4%. The numerical results clearly show that the performance of our proposed method outperforms in comparison with several existing methods with greater accuracy.

Person re-identification (ReID) is an imperative area of pedestrian analysis and has practical applications in visual surveillance. In the person ReID, the robust feature representation is a key issue because of inconsistent visual appearances of a person. Also, an exhaustive gallery search is required to find the target image against each probe image. To answer such issues, this manuscript presents a framework named features-based clustering and deep features in person ReID. The proposed framework initially extracts three types of handcrafted features on the input images including shape, color, and texture for feature representation. To acquire optimal features, a feature fusion and selection technique is applied to these handcrafted features. Afterward, to optimize the gallery search, features-based clustering is performed for splitting the whole gallery into $$k$$ consensus clusters. For relationship learning of gallery features and related labels of the chosen clusters, radial basis kernel is employed. Later on, cluster-wise, images are selected and provided to the deep convolution neural network model to obtain deep features. Then, a cluster-wise feature vector is obtained by fusing the deep and handcrafted features. It follows the feature matching process where multi-class support vector machine is applied to choose the related cluster. Finally, to find accurate matching pair from the classified cluster(s) instead of the whole gallery search, a cross-bin histogram-based distance similarity measure is used. The recognition rate at rank 1 is attained as 46.82%, 48.12%, and 40.67% on selected datasets VIPeR, CUHK01, and iLIDS-VID, respectively. It confirms the proposed framework outperforms the existing ReID approaches.

The most important intricacy when processing natural scene text images is the existence of fog, smoke or haze. These intrusion elements decrease the contrast and disrupt the color fidelity of the image for various computer vision applications. In this paper, such a challenging issue is addressed. The intended work presents a novel method, that is, single image dehazing, based on transmission map. The contributions are performed in the following ways: (1) text extraction from hazy image is not straightforward due to lack of haze-free images and hazy images. To address this limitation, we introduce synthetic natural scene text image composed of pairs of synthetic hazy and corresponding haze-free images using mainstream datasets. Different from existing dehazing datasets, text in hazy images is considered compulsory content, which needs to be separated from background using the recovered image. For doing this, based on transmission map the scenic depth is calculated using haze density and color attenuation to generate depth map. In the next step, raw transmission map is computed, which is further refined using bilateral filtering to preserve edges and avoid possible noise; (2) text region proposals are estimated on the restored images using novel low-level connected component technique and character bounding is employed to complete the process. Finally, the experimentations are carried out on the images selected from standard datasets including MSRA-TD500, SVT and KAIST. The experimental outcomes demonstrate that the intended method performs better when compared with benchmark standard techniques and publically available dehazing datasets.

Electrocardiography is the primary diagnostic tool for measuring the malfunction of different heart activities in the form of various cardiac diseases. Some cardiac diseases require special attention due to the urgency and risk factors involved. Myocardial infarction (MI) is one of the cardiac diseases that require robust identification. Early prediction in MI cases without prior history remains to be an ongoing challenge. This article delivers a major novel contribution in the context of predictive classification of flattened T-wave MI cases. Therefore, a novel model-driven deep deterministic learning (MDDDL) approach is proposed. In MDDDL, two different data sets are used for the execution of operational activities in terms of flattened T-wave predictive classification. The first data set is the publicly available Physikalisch-Technische Bundesanstalt (PTB), and the second data set is exclusively obtained from the University of Malaya Medical Centre (UMMC). Firstly, the systematic behaviour of MDDDL is defined in terms of pattern recognition of extracted features between T-wave alternans and flattened T-wave subjects, and then both data sets are merged considering data fusion approach and pre-defined conditions. Afterwards, the empirical approach is adopted in MDDDL evaluation in relation to global acceptance and state-of-the-art comparison. Finally, some qualitative improvements, such as inclusion of a backtracking factor for rapid prediction of flattened anomalies and increasing the number of features along with enhancement of fusion processes to reduce complexity, are required by the MDDDL and should be covered in future works.

Due to the rapid growth in artificial intelligence (AI) and deep learning (DL) approaches, the security and robustness of the deployed algorithms need to be guaranteed. The security susceptibility of the DL algorithms to adversarial examples has been widely acknowledged. The artificially created examples will lead to different instances negatively identified by the DL models that are humanly considered benign. Practical application in actual physical scenarios with adversarial threats shows their features. Thus, adversarial attacks and defense, including machine learning and its reliability, have drawn growing interest and, in recent years, has been a hot topic of research. We introduce a framework that provides a defensive model against the adversarial speckle-noise attack, the adversarial training, and a feature fusion strategy, which preserves the classification with correct labelling. We evaluate and analyze the adversarial attacks and defenses on the retinal fundus images for the Diabetic Retinopathy recognition problem, which is considered a state-of-the-art endeavor. Results obtained on the retinal fundus images, which are prone to adversarial attacks, are 99% accurate and prove that the proposed defensive model is robust.

Traffic Sign Recognition (TSR) is a crucial step for automated vehicles and driver assistance systems. Automated TSD in an extreme environment has always been challenging due to foggy, rainy, blurry, and cropping images. A real-time TSD model named improved Extreme Learning Machine Network (iELMNet) is proposed to tackle these challenges. Primary modules of iELMNet include: a) Custom DensNet; b) Accurate Anchor Prediction Model (A2PM); c) Scale Transformation (ST), and d) Extreme Learning Machine (ELM) classifier. Convolutional Neural Network (CNN) model improvises edges of traffic-signs using mapped images extracted from handcrafted features. A2PM removes the redundant features to improve efficiency. ST is utilized to allow the proposed technique for detecting these signs of variant sizes. ELM classifier tries to classify traffic signs robustly by minimizing the feature dimensions. The proposed model is evaluated over three publicly available datasets, i.e., CURE-TSR, TT100k, and GTSRB, and acquired 98.63%, 95.22%, and 99.45% precision, respectively. The output of proposed model demonstrates its competence and ability to implement it in a practical environment.

Video based Human Action Recognition (HAR) is an active research field of Machine Learning (ML) and human detection in videos is the most important step in action recognition. Recently, several techniques and algorithms have been proposed to increase the accuracy of HAR process, but margin of improvement still exists. Detection and classification of human actions is a challenging task due to random changes in human appearance, clothes, illumination, and background. In this article, an efficient technique to classify human actions by utilizing steps like removing redundant frames from videos, extracting Segments of Interest (SoIs), feature descriptor mining through Geodesic Distance (GD), 3D Cartesian-plane Features (3D-CF), Joints MOCAP (JMOCAP) and n-way Point Trajectory Generation (nPTG). A Neuro Fuzzy Classifier (NFC) is used at the end for the classification purpose. The proposed technique is tested on two publicly available datasets including HMDB-51 and Hollywood2, and achieved an accuracy of 82.55% and 91.99% respectively. These efficient results prove the validity of proposed model.

Documents are stored in a digital form across several organizations. Printing this amount of data and placing it into folders instead of storing digitally is against the practical, economical, and ecological perspective. An efficient way of retrieving data from digitally stored documents is also required. This article presents a real-time supervised learning technique for document classification based on deep convolutional neural network (DCNN), which aims to reduce the impact of adverse document image issues such as signatures, marks, logo, and handwritten notes. The proposed technique&rsquo, s major steps include data augmentation, feature extraction using pre-trained neural network models, feature fusion, and feature selection. We propose a novel data augmentation technique, which normalizes the imbalanced dataset using the secondary dataset RVL-CDIP. The DCNN features are extracted using the VGG19 and AlexNet networks. The extracted features are fused, and the fused feature vector is optimized by applying a Pearson correlation coefficient-based technique to select the optimized features while removing the redundant features. The proposed technique is tested on the Tobacco3482 dataset, which gives a classification accuracy of 93.1% using a cubic support vector machine classifier, proving the validity of the proposed technique.

As the number of internet users increases so does the number of malicious attacks using malware. The detection of malicious code is becoming critical, and the existing approaches need to be improved. Here, we propose a feature fusion method to combine the features extracted from pre-trained AlexNet and Inception-v3 deep neural networks with features attained using segmentation-based fractal texture analysis (SFTA) of images representing the malware code. In this work, we use distinctive pre-trained models (AlexNet and Inception-V3) for feature extraction. The purpose of deep convolutional neural network (CNN) feature extraction from two models is to improve the malware classifier accuracy, because both models have characteristics and qualities to extract different features. This technique produces a fusion of features to build a multimodal representation of malicious code that can be used to classify the grayscale images, separating the malware into 25 malware classes. The features that are extracted from malware images are then classified using different variants of support vector machine (SVM), k-nearest neighbor (KNN), decision tree (DT), and other classifiers. To improve the classification results, we also adopted data augmentation based on affine image transforms. The presented method is evaluated on a Malimg malware image dataset, achieving an accuracy of 99.3%, which makes it the best among the competing approaches.

The citrus plants such as lemons, mandarins, oranges, tangerines, grapefruits, and limes are commonly grown fruits all over the world. The citrus producing companies create a large amount of waste every year whereby 50% of citrus peel is destroyed every year due to different plant diseases. This paper presents a survey on the different methods relevant to citrus plants leaves diseases detection and the classification. The article presents a detailed taxonomy of citrus leaf diseases. Initially, the challenges of each step are discussed in detail, which affects the detection and classification accuracy. In addition, a thorough literature review of automated disease detection and classification methods is presented. To this end, we study different image preprocessing, segmentation, feature extraction, features selection, and classification methods. In addition, also discuss the importance of features extraction and deep learning methods. The survey presents the detailed discussion on studies, outlines their strengths and limitations, and uncovers further research issues. The survey results reveal that the adoption of automated detection and classification methods for citrus plants diseases is still in its infancy. Hence new tools are needed to fully automate the detection and classification processes.

Image based text extraction is a popular and challenging research field in computer vision in recent times. In this paper, an exigent aspect such as natural scene text identification and extraction has been investigated due to cluttered background, unstructured scenes, orientations, ambiguities and much more. For text identification, contrast enhancement is done by applying LUV channel on an input image to get perfect stable regions. Then L-Channel is selected for region segmentation using standard segmentation technique MSER. In order to differentiate among text/non-text regions, various geometrical properties are also considered in this work. Further, classification of connected components is performed to obtain segmented image by the fusion of two feature descriptors LBP and T-HOG. Firstly both features descriptors are separately classified using linear SVM (s) . Secondly the results of both are combined by applying weighted sum fusion technique to classify into text/non-text portions. In text recognition, text regions are recognized and labeled with a novel CNN network. The CNN output is stored in a text file to make a text word. Finally, the text file is searched through lexicon for proper optimized scene text word incorporating hamming distance (error correction) technique if necessary.

Background: Breast cancer is considered as one of the most perilous sickness among females worldwide and the ratio of new cases is increasing yearly. Many researchers have proposed efficient algorithms to diagnose breast cancer at early stages, which have increased the efficiency and performance by utilizing the learned features of gold standard histopathological images. Objective: Most of these systems have either used traditional handcrafted or deep features, which had a lot of noise and redundancy, and ultimately decrease the performance of the system. Methods: A hybrid approach is proposed by fusing and optimizing the properties of handcrafted and deep features to classify the breast cancer images. HOG and LBP features are serially fused with pre-trained models VGG19 and InceptionV3. PCR and ICR are used to evaluate the classification performance of the proposed method. Results: The method concentrates on histopathological images to classify the breast cancer. The performance is compared with the state-of-the-art techniques, where an overall patient-level accuracy of 97.2% and image-level accuracy of 96.7% is recorded. Conclusion: The proposed hybrid method achieves the best performance as compared to previous methods and it can be used for the intelligent healthcare systems and early breast cancer detection.

During recent years, the demand for automated video surveillance has become the subject of much attention. Currently, video surveillance is an essential part of security and monitoring in banks, streets, buildings, department stores, stadiums, highways, railway stations, and crowded gatherings. One important reason for using video surveillance is to counter terrorism. In this context, a multi-camera handoff system for person re-identification using a color-based global appearance model is presented. We address two main factors that affect re-identification performance in this paper: (1) robust features (viewpoint orientation and indoor/outdoor cinematography) and (2) re-identification or classification. A hexagonal-superpixel method is introduced to minimize the false recognition rate due to the above-mentioned challenges. First we detect the human or object in a video and use a mixture of Gaussian models extended with a trajectory-learning algorithm to solve the trajectory dilemma. Second, a novel feature descriptor is proposed in which sensitive features are extracted using the proposed Color Hexagonal-SIFT and Color Histogram Features (CHF) methods. Third, we generate a combined appearance feature descriptor for re-identifying the human in different scenes. Finally, we introduce a Time-Tree-based learning method to minimize the gallery set׳s overhead and increase the accuracy of the re-identification results. The proposed framework is tested on videos taken using three cameras placed in three different locations. We also tested this framework on publically available data from the i-LIDS and VIPeR datasets. The results show outstanding speed and re-identification accuracy in dealing with the above-mentioned issues.

Provocative heart disease is related to ventricular arrhythmias (VA). Ventricular tachyarrhythmia is an irregular and fast heart rhythm that emerges from inappropriate electrical impulses in the ventricles of the heart. Different types of arrhythmias are associated with different patterns, which can be identified. An electrocardiogram (ECG) is the major analytical tool used to interpret and record ECG signals. ECG signals are nonlinear and difficult to interpret and analyze. We propose a new deep learning approach for the detection of VA. Initially, the ECG signals are transformed into images that have not been done before. Later, these images are normalized and utilized to train the AlexNet, VGG-16 and Inception-v3 deep learning models. Transfer learning is performed to train a model and extract the deep features from different output layers. After that, the features are fused by a concatenation approach, and the best features are selected using a heuristic entropy calculation approach. Finally, supervised learning classifiers are utilized for final feature classification. The results are evaluated on the MIT-BIH dataset and achieved an accuracy of 97.6% (using Cubic Support Vector Machine as a final stage classifier).

Plant diseases are accountable for economic losses in an agricultural country. The manual process of plant diseases diagnosis is a key challenge from last one decade; therefore, researchers in this area introduced automated systems. In this research work, automated system is proposed for citrus fruit diseases recognition using computer vision technique. The proposed method incorporates five fundamental steps such as preprocessing, disease segmentation, feature extraction and reduction, fusion, and classification. The noise is being removed followed by a contrast stretching procedure in the very first phase. Later, watershed method is applied to excerpt the infectious regions. The shape, texture, and color features are subsequently computed from these infection regions. In the fourth step, reduced features are fused using serial-based approach followed by a final step of classification using multiclass support vector machine. For dimensionality reduction, principal component analysis is utilized, which is a statistical procedure that enforces an orthogonal transformation on a set of observations. Three different image data sets (Citrus Image Gallery, Plant Village, and self-collected) are combined in this research to achieving a classification accuracy of 95.5%. From the stats, it is quite clear that our proposed method outperforms several existing methods with greater precision and accuracy.

Automated Pedestrian classification has been the topic of interest over past many years. Various traditional and neural network based approaches have been adopted to recognize pedestrians. In this paper, we employ stacked sparse autoencoder as a deep learning building block for object feature extraction. Salient feature maps of the input images are generated with the help of SLIC superpixel-based graph manifold ranking. These salient maps are then served as input to the stacked sparse autoencoder in the next step. Softmax regression is selected as the classifier to identify the pedestrians. The learning and testing procedures are performed with the publically available datasets. The experiments results obtained are compared with a recent traditional pedestrian method. The better performance rates depict the robustness of the proposed approach.

Efficient and fast segmentation of fruit symptoms is one of the major businesses nowadays in the agro-economy. Manual segmentation and recognition of fruit symptoms is a hard job because of many aspects such like time-consuming and waste of money. Lately, several scholars came up with image processing and pattern recognition based methods for segmentation and recognition of fruit symptoms based on their features such as color, texture, and shape. In this article, we proposed an automated system for segmentation and recognition of grape leaf diseases. The proposed system comprises of four main steps. In first step, a local contrast haze reduction (LCHR) enhancement technique is proposed for increasing the local contrast of symptoms. Thereafter, LAB color transformation is held in the second step and the best channel is selected based on the pixels information that is later utilized into thresholding function. Color, texture, and geometric features are extracted and fused by canonical correlation analysis (CCA) approach. At the time of features fusion, a noise is added in the form of irrelevant and redundant features that are removed by Neighborhood Component Analysis (NCA). The classification of final reduced features is then performed by M-class SVM. The introduced system is assessed on Plant Village dataset of three types of grape leaf diseases such as black measles, black rot, and leaf blight including healthy. The proposed method acquired an average segmentation accuracy rate of 90% and classification accuracy is above 92% which is superior in contrast of existing techniques.

Advanced laboratory technology has made blood testing more automated, robust and tests are being implemented comprehensively. Leukocytes type differentiation is a critical hematological images analysis step of blood film as it delivers valuable information in diagnosis of several diseases. At present, the morphological examination of leukocytes is done manually and this process is very tedious, inefficient and slow. Although many white blood cells detection or classification techniques are presented by different researchers, there is still a need of fully automated and an efficient detection system of blood cells with its particular types for an early diagnosis of leukemia. This paper presents a technique for the classification of protuberant types of leukocytes and early diagnosis of leukemia. The work is divided into the following main stages: (a) image augmentation, (b) wavelet composition and decomposition for attaining high and low frequency bands of the cell image, (c) convolutional neural network (CNN) training model for the classification of leukocytes categories and (d) prediction of leukemia. The main intention behind this study is to develop an automated, robust and efficient classification and detection system of leukocytes for microscopic blood images.

In computer vision applications such as person re-identification the optimization of rank list is an important issue. In order to address this issue, a multi-feature fusion based re-ranking technique is proposed. In most of the conventional methods, a long feature vector is formulated from a single modality. Whereas, in the proposed approach, multiple features from the image are extracted and combined into a unified/hybrid vector. Later a joint feature vector is presented after fusion. The Mahalanobis distance is calculated for checking the similarity between the image pairs. A tree based re-ranking algorithm is also proposed that used this combined feature vector and the distance metric. Therefore, by effective use of each feature type, better re-rank can be achieved. We assessed the proposed method on publically available datasets VIPeR and ETHZ. Experimental results demonstrate that the presented approach performs well than exploiting each individual feature.

In the area of medical imaging and computer vision, automatic diagnosis of ulcer and bleeding from wireless capsule endoscopy images has been an active research domain. It contains several challenges including low contrast, complex background, lesion shape and color which affect its segmentation and classification accuracy. In this article, a novel method for automated detection and classification of stomach infection is implemented. The proposed method consists of four major steps including preprocessing, lesion segmentation, image representation and classification. The lesion contrast is improved in preprocessing step by employing 3D-box filtering, 3D-median filtering and HSV transformation. In the second step, geometric features are extracted and applied to the saturated channel to give a binary image. The binary image is further improved by fusion of generated mask. After that, extraction of three types of features including color, shape and surf is performed from HSV and binary segmented images and their information is fused by a serial based method. A principal component analysis (PCA) and correlation coefficient based feature selection approach is proposed which are classified by multi class support vector machine (M-SVM). The proposed method is evaluated on personally collected images of three different classes including ulcer, bleeding and healthy. The M-SVM performs well with a maximum accuracy of 98.3% which shows the authenticity of presented method.

Currently, identifying humans using biomechanics-based approaches has gained a lot of significance for person re-identification. Biomechanics-based approaches use knee-hip angle–angle relationships and body movements for person re-identification. Generally, biomechanics of human walking and running is used for person re-identification. In fact, person re-identification is a complex and important task in academia as well as industry and remains an unsolved issue in the computer vision field. The subjects most commonly addressed regarding person re-identification include significant feature extraction that can function accurately with invariant appearance and robust classification. In this study, a significant color feature descriptor is proposed by combining dense color-SIFT and global convex hull salience region features. First convex hull boundary points are detected using the SIFT technique. Furthermore, it is extended with Grubb’s outlier test to eliminate the outlier points detected by SIFT and mark the saliency region via convex hull. Then dense-SIFT and dense-CHF methods are used to extract local and global features within the convex hull region, respectively. Finally, the pre-ranked common nearest neighbor selection technique is applied to minimize overhead of dataset and generate more robust rank classification. The proposed technique is tested using three-camera database video sequences and three publicly available datasets, namely i-LIDS, VIPeR and GRID. Performance of re-identification system is evaluated using a statistical method with CMC curves. The results show better re-identification accuracy in solving the aforementioned problems.