Classifier Adaptive Fusion: Deep Learning for Robust Outdoor Vehicle Visual Tracking

Robust visual tracking for outdoor vehicle is still a challenging problem due to large appearance variations caused by illumination variation, occlusion and scale variation, etc. In this paper, a deep-learning-based approach for robust outdoor vehicle tracking is proposed. Firstly, a stacked denoising auto-encoder is pre-trained to learn the feature representation way of images. Then, a k-sparse constraint is added to the stacked denoising auto-encoder and the encoder of k-sparse stacked denoising auto-encoder (kSSDAE) is connected with a classification layer to construct a classification neural network. After fine-tuning, the classification neural network is applied to online tracking under particle filter framework. Extensive tracking experiments are conducted on a challenging single object online tracking evaluation platform benchmark to verify the effectiveness of our tracker. Experiments show that our tracker outperforms most state-of-the-art trackers.

Human Activity Recognition (HAR) by employing smart home sensors is a topic that is undergoing intense research in the area of ambient-supported living and the basis of ubiquitous computing in smart environments. Recently, HAR majorly utilizes Deep Learning (DL) method since they employ representation learning methods that could automatically produce maximum factors from raw input datasets resulting from the sensor in the absence of human interference and could detect Hidden Layer (HL) patterns in a dataset. The study presents an Improved Manta-Ray Foraging Optimization with Deep Autoencoder (IMRFO-DAE) model for HAR. The presented algorithm majorly recognizes the diverse types of human activities. The presented IMRFO-DAE model pre-processes the human activity data via a standardization approach to obtain this. Next, the IMRFO-DAE method uses the DAE model to perform the Activity Recognition (AR) process. To advance the AR accomplishment of the DAE model, the IMRFO method is applied as a hyperparameter optimizer. Moreover, the IMRFO model is derived by modifying the initialization of the MRFO model by implementing the chaotic concept. An extensive range of simulations was performed to depict the enhanced efficiency of the IMRFO-DAE approach. The simulation results assured the improved outcomes of the IMRFO-DAE approach compared to existing approaches.

Melanoma, a kind of fatal skin cancer, originates in melanin secreting cells of the dermis. Disease identification in the early stages assures a high survival rate for the patient. Most of the existing techniques retard the cancer detection phase. Surface-Enhanced Raman Spectroscopy (SERS) can capture fine details from the specimens that machine learning models can utilize to discriminate between healthy and diseased individuals rapidly. Our research work proposes a deep autoencoder based hybrid dimensionality reduction approach with a machine learning model on SERS spectrums of human skin fibroblast for melanoma cancer diagnostics. SERS measurements of 307 samples in total, belonging to two different classes, such as normal (157 samples) and malignant melanoma (150 samples), are used in this study. The SERS spectra measurements for both the samples lie between 100cm-1 and 4278cm-1. The variations in the intensity of Raman bands between both classes are intrinsically subtle. Neighborhood Component Analysis (NCA) technique has been exerted to transform 2090 dimensional spectral features into 2090 dimensional vectors and then the Deep Autoencoder (DAE) model is used to handle the nonlinearity in the data and produce the latent space, while Linear Discriminant Analysis (LDA) classifier have been employed for discriminating the normal and cancer cells. The k-fold cross-validation technique with a k value of 10 is implemented to assess the metrics of the model. The stated hybrid (NCA and DAE) model with 10-dimension latent space achieves an accuracy of 98%, the sensitivity of 99% and specificity of 97%, respectively. Due to the high-intensity nature of the SERS spectrum, the existing linear dimensionality reduction based discriminating model fails if the class label (Normal or Cancer) gets distributed on the low variance side. The proposed methodology captures both linear and nonlinear underlying structures present in the spectrums, resulting in better classification compared to the standard dimensionality reduction techniques.

Object slip perception is essential for mobile manipulation robots to perform manipulation tasks reliably in the dynamic real-world. Traditional approaches to robot arms’ slip perception use tactile or vision sensors. However, mobile robots still have to deal with noise in their sensor signals caused by the robot’s movement in a changing environment. To solve this problem, we present an anomaly detection method that utilizes multisensory data based on a deep autoencoder model. The proposed framework integrates heterogeneous data streams collected from various robot sensors, including RGB and depth cameras, a microphone, and a force-torque sensor. The integrated data is used to train a deep autoencoder to construct latent representations of the multisensory data that indicate the normal status. Anomalies can then be identified by error scores measured by the difference between the trained encoder’s latent values and the latent values of reconstructed input data. In order to evaluate the proposed framework, we conducted an experiment that mimics an object slip by a mobile service robot operating in a real-world environment with diverse household objects and different moving patterns. The experimental results verified that the proposed framework reliably detects anomalies in object slip situations despite various object types and robot behaviors, and visual and auditory noise in the environment.

In visual tracking, developing an efficient appearance model is a challenging task due to the influence of various factors, such as illumination variation, occlusion, background clutter, and so on. Existing tracking algorithms use appearance samples from previous frames to form a template set upon which target appearance models are built. However, these appearance models are data-dependent, so they may be corrupted by significant appearance variation. It is difficult to update the templates in challenging environments. In this paper, we propose a robust visual tracking algorithm with an adaptive appearance model using a point-to-set metric learning technique. To do this, we first model a target representation using a set of target templates and a regularized affine hull (RAH) spanned by the target templates. Then, we learn a point-to-set distance metric, which is incorporated into the optimization process to obtain an adaptive target representation. The RAH model covers unseen target appearances by affine combinations of the target templates. Based on the proposed target appearance model, we design an effective template update scheme by adjusting the weights of the target templates. Experimental results on challenging video sequences with comparisons to several state-of-the-art tracking algorithms demonstrate the effectiveness and robustness of the proposed tracking algorithm.

Sparse representation has been widely applied for modeling object appearance in the tracking process. Various sparsity models are coded by single feature with fixed local patches, which neglects the potential spatial information inside the appearance model. This way also gives a low robust effect on

Over the past decades, Session Initiation Protocol (SIP) has been commonly utilized in signaling protocol to control and signal multimedia communication sessions for providing various services to users. Video calls and internet telephony over Internet Protocol (IP) are the most common applications on SIP. Generally, the SIP acts as an unsafe communication platform, due to the exchange of information via public channels. Thus, the security of the SIP network should be increased prominently and require secure authentication and key establishment schemes. In this article, a deep learning model fractional lyrebird optimization algorithm–deep stack autoencoder (FLOA‐DSA) is developed for the detection of SIP‐based distributed reflection denial‐of‐service (DRDoS) attacks in SIP network. Here, the authentication process is carried out under different phases based on various mathematical model functionalities. Moreover, the deep stack autoencoder (DSA) model is used to perform intrusion detection in the attack validation phase. The detection performance of DSA is increased by training using the fractional lyrebird optimization algorithm (FLOA); FLOA provides a more effective, reliable, and scalable optimization strategy for training DSAs than traditional algorithms. Its ability to dynamically interact with DSA’s learning dynamics makes it not just a suitable choice, but the optimal optimization choice for this hybrid deep learning system. Further, the detection performance of the FLOA‐DSA is evaluated based on the metrics, such as detection rate, F ‐measure, false positive rate (FPR), memory usage, validation time, conditional privacy, response time, energy efficiency, and latency, and the FLOA‐DSA obtained best results of 90.12%, 91.40%, 9.13%, 6.9 MB, 48.5 s, 3.86 s, 88.88 bits/J, and 2.68 s. The F ‐measure of the proposed method is 9.06%, 6.63%, 4.80%, and 2.61% improved than other existing comparative methods.

In this paper, we propose a robust visual tracking method by casting tracking as a sparse approximation problem in a particle filter framework. In this framework, occlusion, noise, and other challenging issues are addressed seamlessly through a set of trivial templates. Specifically, to find the tracking target in a new frame, each target candidate is sparsely represented in the space spanned by target templates and trivial templates. The sparsity is achieved by solving an l1-regularized least-squares problem. Then, the candidate with the smallest projection error is taken as the tracking target. After that, tracking is continued using a Bayesian state inference framework. Two strategies are used to further improve the tracking performance. First, target templates are dynamically updated to capture appearance changes. Second, nonnegativity constraints are enforced to filter out clutter which negatively resembles tracking targets. We test the proposed approach on numerous sequences involving different types of challenges, including occlusion and variations in illumination, scale, and pose. The proposed approach demonstrates excellent performance in comparison with previously proposed trackers. We also extend the method for simultaneous tracking and recognition by introducing a static template set which stores target images from different classes. The recognition result at each frame is propagated to produce the final result for the whole video. The approach is validated on a vehicle tracking and classification task using outdoor infrared video sequences.

In visual tracking, developing a robust appearance model is a challenging task due to variations of object appearances such as background clutter, illumination variation and partial occlusion. In existing tracking algorithms, a target candidate is represented by linear combinations of target templates. However, the relationship between a target candidate and the corresponding target templates is nonlinear because of appearance variations. In this paper, we propose a kernelized convex hull based target representation for visual tracking. Namely, a target is represented by a nonlinear combination of target templates in a mapped higher dimensional feature space. The convex hull model can covers the target appearances that do not appear in the target templates. Experimental results demonstrate the robustness and effectiveness of the proposed tracking algorithm against several state-of-the-art tracking algorithms.

Face recognition is one of the biometric identification methods with the highest potential. The existing face recognition algorithms relying on the texture information of face images are affected greatly by the variation of expression, scale and illumination. Whereas the algorithms based on the shape topology weaken the influence of illumination to some extent, but the impact of expression, scale and illumination on face recognition is still unsolved. To this end, we propose a new method for face recognition by integrating texture information with shape information, called biview face recognition algorithm. The texture models are constructed by using subspace learning methods and shape topologies are formed by building graphs for face images. The proposed biview face recognition method is compared with recognition algorithms merely based on texture or shape information. Experimental results of recognizing faces under the variation of illumination, expression and scale demonstrate that the performance of the proposed biview face recognition outperforms texture-based and shape-based algorithms.

In real‐time, visually impaired persons face challenging issues in identifying and avoiding obstacles in their path, making it difficult to move around freely and confidently. Object detection technology benefits visually impaired persons and contains the procedure of determining several individual objects from the image and recognizing their place with bounding boxes with class labels. Both exploit typical machine learning (ML) approaches and utilize deep learning (DL) models to execute object detection methods. Recently, several researchers have leveraged DL techniques such as convolutional neural networks (CNNs) for solving object detection problems and presented various recent object detection approaches. The study presents a sine cosine optimization with DL–driven object detector (SCODL–ODC) technique for visually impaired persons. The presented SCODL–ODC technique detects and classifies the objects, which can help them navigate their environment more efficiently and avoid obstacles. To accomplish this, the proposed SCODL–ODC technique employs a YOLOv5 object detector at the initial stage to detect the objects accurately, and the SCO model performs its hyperparameter tuning process. The following step is to classify them using the deep stacked autoencoder (DSAE) model when the objects are detected. Finally, the Harris Hawks optimization (HHO) technique is used to adjust the hyperparameters of the DSAE model. The experimentation outcome of the SCODL–ODC method is examined on the object detection dataset. The investigational validation of the SCODL–ODC method portrayed a superior accuracy value of 99.31% over existing models.

The proposed work introduces a dual-stage deep capsule autoencoder (DSDC-AE) model for fake news detection on Twitter data. Initially, the input Twitter data are pre-processed using tokenisation, stemming, stop word removal and lemmatisation. The text features are extracted from the pre-processed data using Improved Term Frequency Inverse Document Frequency (ITF-IDF), Unigrams, Bigrams, Enhanced Bag of Words (EBoW) and Advanced Word2vec (Word2 Vector). To minimise the large feature dimensionality, the proposed work uses the Horse Herd Optimisation algorithm (HOA) for the feature selection. Finally, the selected features are subjected to the proposed classifier, in which the fake news from Twitter data is detected and classifies the given input data as real or fake. The proposed DSDC-AE model uses three datasets for the analysis: fake news detection on Twitter EDA, ISOT dataset and FakeNewsNet dataset. The mentioned dataset obtains the accuracy of 99.52%, 99.51% and 99.47%.

Consumer-oriented sensory optimization of yogurt: An artificial intelligence approach

Treatment resistance prevents patients with preoperative chemoradiotherapy or targeted radiolabeled immunotherapy from achieving a good result, which remains a major challenge in the prostate cancer (PCa) area. A novel integrative framework combining a machine learning workflow with proteogenomic profiling was used to identify predictive ultrasound biomarkers and classify patient response to radiolabeled immunotherapy in high-risk PCa patients who are treatment resistant. The deep stacked autoencoder (DSAE) model, combined with Extreme Gradient Boosting, was designed for feature refinement and classification. The Cancer Genome Atlas and an independent radiotherapy-treated cohort have been utilized to collect multiomics data through their respective applications. In addition to genetic mutations (whole-exome sequencing), these data contained proteomic (mass spectrometry) and transcriptomic (RNA sequencing) data. Maintaining biological variety across omics layers while reducing the dimensionality of the data requires the use of the DSAE architecture. Resistance phenotypes show a notable relationship with proteogenomic profiles, including DNA repair pathways (Breast Cancer gene 2 [BRCA2], ataxia-telangiectasia mutated [ATM]), androgen receptor (AR) signaling regulators, and metabolic enzymes (ATP citrate lyase [ACLY], isocitrate dehydrogenase 1 [IDH1]). A specific panel of ultrasound biomarkers has been confirmed in a state deemed preclinical using patient-derived xenografts. To support clinical translation, real-time phenotypic features from ultrasound imaging (e.g., perfusion, stiffness) were also considered, providing complementary insights into the tumor microenvironment and treatment responsiveness. This approach provides an integrated platform that offers a clinically actionable foundation for the development of radiolabeled immunotherapy drugs before surgical operations.

By integrating hierarchical feature modeling of auditory information using deep neural networks (DNNs), recent functional magnetic resonance imaging (fMRI) encoding studies have revealed the hierarchical neural auditory representation in the superior temporal gyrus (STG). Most of these studies adopted supervised DNNs (e.g., for audio classification) to derive the hierarchical feature representation of external auditory stimuli. One possible limitation is that the extracted features could be biased toward discriminative features while ignoring general attributes shared by auditory information in multiple categories. Consequently, the hierarchy of neural acoustic processing revealed by the encoding model might be biased toward classification. In this study, we explored the hierarchical neural auditory representation via an fMRI encoding framework in which an unsupervised deep convolutional auto-encoder (DCAE) model was adopted to derive the hierarchical feature representations of the stimuli (naturalistic auditory excerpts in different categories) in fMRI acquisition. The experimental results showed that the neural representation of hierarchical auditory features is not limited to previously reported STG, but also involves the bilateral insula, ventral visual cortex, and thalamus. The current study may provide complementary evidence to understand the hierarchical auditory processing in the human brain.