Visual Object Multimodality Tracking Based on Correlation Filters for Edge Computing

Thermal infrared and visible sequences fusion tracking based on a hybrid tracking framework with adaptive weighting scheme

Multiple correlation filters with gaussian constraint for fast online tracking

Aiming at the problem of severe accuracy degradation in the traditional KCF (Kernel correlation filter) target tracking algorithm in a complex environment, we propose an improved kernel correlation filter target tracking algorithm. First, the target scale is estimated by fusing the scale filter in the DSST (Discriminative Scale Space Tracking) algorithm to solve the traditional KCF algorithm cannot estimate the scale. Secondly, So as to solve the KCF algorithm's limited ability to express the features of the target appearance model, by fusing HOG (Histogram of Oriented Gradient) features and CN (Color Name) features, the algorithm's ability to extract target features is enhanced. Finally, by introducing peak response fluctuation detection, the credibility of the tracking target is judged, so as to decide whether to update the filter model. The experiment tested the overall tracking performance of the improved KCF algorithm by using the video sequences in the OTB2013 and OTB2015 standard data sets. The experimental results show that the accuracy and performance of the improved algorithm in complex environments have been significantly improved.

Due to the inherent characteristics of synthetic aperture radar (SAR) imaging, SAR ship features are not obvious and the category distribution is unbalanced, which makes the task of ship recognition in SAR images quite challenging. To address the above problems, a two-level feature-fusion ship recognition strategy combining the histogram of oriented gradients (HOG) features with the dual-polarized data in the SAR images is proposed. The proposed strategy comprehensively utilizes the features extracted by the HOG operator and the shallow and deep features extracted by the Siamese network in the dual-polarized SAR ship images, which can increase the amount of information for the model learning. First, the Siamese network is used to extract the shallow and deep features from the dual-polarized SAR images, and then the HOG feature of the dual-polarized SAR images is also extracted. Furthermore, the bilinear transformation layer is used for fusing the HOG features from dual-polarized SAR images, and the grouping bilinear pooling process is used for fusing the dual-polarized shallow feature and deep feature extracted by the Siamese network, respectively. Finally, the catenation operation is used for fusing the dual-polarized HOG features and dual-polarized shallow feature and deep feature, respectively, which are used for the recognition of the SAR ship targets. Experimental results tested on the OpenSARShip2.0 dataset demonstrate the correctness and effectiveness of the proposed strategy, which can effectively improve the recognition performance of the ship targets by fusing the different level features of the dual-polarized SAR images.

Handwritten signature verification is a critical aspect of personal authentication in numerous sectors, including banking, legal, and access control systems. This paper provides a comprehensive investigation into writer-dependent signature verification techniques, spanning from traditional approaches such as Histogram of Oriented Gradients (HOG) and Scale-Invariant Feature Transform (SIFT) to cutting-edge deep learning architectures like Siamese and triplet networks. Through extensive experimentation and meticulous analysis conducted on benchmark datasets, we rigorously compare and evaluate the efficacy of these approaches. Our study reveals intriguing insights into the performance characteristics of different methodologies, shedding light on their strengths, weaknesses, and areas for improvement. Notably, we observe that while traditional feature-based methods excel in capturing discriminative information from signature images, deep learning architectures offer enhanced flexibility and adaptability, particularly in scenarios with large and diverse datasets. By combining the strengths of both paradigms, we demonstrate the potential for building robust and reliable signature verification systems capable of handling real-world challenges and variations. We highlight the importance of carefully curated datasets and well-designed network architectures in achieving optimal results, thereby providing valuable guidelines for practitioners and researchers in the field. In conclusion, this study contributes significant insights to the advancement of handwritten signature verification systems, emphasizing the importance of leveraging both traditional methodologies and modern deep learning techniques for building secure and dependable authentication mechanisms. By addressing key challenges and exploring novel approaches, our research aims to foster innovation and enhance the security measures associated with handwritten signature authentication in diverse applications. Keywords—Handwritten signature verification, Writer-dependent approach, Feature-based methods, Deep learning, Siamese networks, Triplet networks, Benchmark datasets, Performance evaluation, Authentication mechanisms, Security measures.

Objectives: To present an automated computer vision-based traffic sign detection model created to identify roadside signs accurately in different environmental conditions. The goal is to improve detection accuracy using a lightweight architecture suitable for mobile devices. Methods: This study presents CV-TSD model. The technique integrates Histogram of Oriented Gradients (HoG), Local Binary Pattern (LBP), and TensorFlow-based machine learning for image recognition. Traffic sign datasets are collected from various online fora, annotated, and trained using convolutional neural networks. The proposed system uses Histogram of Oriented Gradients (HoG), Local Binary Pattern (LBP), and deep learning with TensorFlow for effective image classification. A custom dataset of traffic sign images was collected from several online sources, manually labelled, and used to retrain a convolutional neural network using transfer learning. Three environmental categories—dark shade, cloudy, and rainy—were used to test classification robustness. We assessed performance with training accuracy, validation accuracy, and cross-entropy loss. Findings: Experimental results show that CV-TSD consistently outperforms traditional HoG and LBP methods across all metrics. The system achieves a maximum training accuracy of 98% and a validation accuracy of 97% with a limited number of image samples. Furthermore, the model keeps stable cross-entropy values, indicating effective learning without overfitting. The lightweight design allows reliable performance even with smaller datasets, making it ideal for mobile and intelligent transportation systems. Novelty: Unlike standard traffic sign detection methods, CV-TSD mimics human visual perception by merging semantic image understanding with deep learning. Its ability to classify accurately with minimal data and low computational demands marks a significant advancement for real-time applications. Keywords: Traffic Sign Detection, Deep Learning, Image Classification, TensorFlow, HoG, LBP, Transfer Learning

Face recognition is a process of verifying an individual using facial images and it is widely employed in identifying people on social media platforms, validating identity at ATMs, finding missing persons, controlling access to sensitive areas, finding lost pets, etc. Face recognition is still a trending research area because of various challenges like illumination variations, different poses, and expressions of the person. Here, a novel methodology is introduced for face recognition using Histogram of Oriented Gradients (HOG), histogram of Local Binary Patterns (LBP), and Convolutional Neural Network (CNN). The features from HOG, histogram of LBP, and deep features from the proposed CNN are linearly concatenated to produce the feature space and then classified by Support Vector Machine. The face databases ORL, Extended Yale B, and CMUPIE are used for experimental work and attained a recognition rate of 98.48%, 97.33%, and 97.28% respectively.

In recent years, with the development of the Internet of Things (IoT) technology, a large amount of data can be captured from sensors for real-time analysis. By monitoring the traffic video data from the IoT, we can detect the anomalies that may occur and evaluate the security. However, the number of traffic anomalies is extremely limited, so there is a severe over-fitting problem when using traditional deep learning methods. In order to solve the problem above, we propose a similarity metric Convolutional Neural Network (CNN) based on a channel attention model for traffic anomaly detection task. The method mainly includes (1) A Siamese network with a hierarchical attention model by word embedding so that it can selectively measure similarities between anomalies and the templates. (2) A deep transfer learning method can automatically annotate an unlabeled set while fine-tuning the network. (3) A background modeling method combining spatial and temporal information for anomaly extraction. Experiments show that the proposed method is three percentage points higher than deep convolutional generative adversarial network (DCGAN) and five percentage points higher than AutoEncoder on the accuracy. No more time consumption is needed for the annotation process. The extracted candidates can be classified correctly through the proposed method.

Computer vision enables a wide range of applications in robotics/drones, self-driving cars, smart Internet of Things, and portable/wearable electronics. For many of these applications, local embedded processing is preferred due to privacy and/or latency concerns. Accordingly, energy-efficient embedded vision hardware delivering real-time and robust performance is crucial. While deep learning is gaining popularity in several computer vision algorithms, a significant energy consumption difference exists compared to traditional hand-crafted approaches. In this paper, we provide an in-depth analysis of the computation, energy and accuracy trade-offs between learned features such as deep Convolutional Neural Networks (CNN) and hand-crafted features such as Histogram of Oriented Gradients (HOG). This analysis is supported by measurements from two chips that implement these algorithms. Our goal is to understand the source of the energy discrepancy between the two approaches and to provide insight about the potential areas where CNNs can be improved and eventually approach the energy-efficiency of HOG while maintaining its outstanding performance accuracy.

In this study, we present the performance of Support Vector Machine (SVM), Convolution Neural Network (CNN), and Artificial Neural Network (ANN) with Bag of Words (BoW), Histogram of Oriented Gradients (HOG), and Image Pixels (IP) for face recognition. SVM, CNN, and ANN are machine learning approaches and has been used for pattern recognition, especially in face recognition technology. BoW, HOG, and IP are being used for image feature extraction. The testing has been conducted from publicly available AT&T face database. Every individual subject consists of 10 images with different facial expression, different illumination and the dimensions of the images is unified as 92-by-112 pixels with PGM formats. SVM achieved recognition accuracy of 97.00%, 96.00%, and 98.00% with subsequently BoW, HOG, and IP. CNN achieved recognition accuracy of 94.00%, 99.00%, and 99.50% with subsequently BoW, HOG, and IP. ANN achieved recognition accuracy of 96.00%, 99.00%, and 99.50% with subsequently BoW, HOG, and IP. The experimental results show that the IP with ANN approach clearly outperformed the others approaches.

Face recognition has emerged as a critical biometric technology for applications in surveillance, authentication, and security. However, variations in illumination, pose, and facial expressions significantly reduce recognition accuracy in practical environments. This research proposes a robust and adaptive framework, DCNN-RRDA, which integrates traditional feature extraction, deep learning, and evolutionary optimization. Facial features are initially captured using Histogram of Oriented Gradients (HOG), followed by hierarchical feature learning through a Deep Convolutional Neural Network (DCNN). Hyperparameters and network weights are optimized using the Refined Red Deer Algorithm (RRDA), which incorporates adaptive, chaotic, and inverse learning strategies to enhance convergence and avoid local optima. The framework was trained and evaluated on a publicly available Kaggle face dataset comprising approximately 16,700 images. Performance was assessed using accuracy, precision, recall, and F1-score, and compared against baseline models including Inception v3, ResNet50, and VGG16. Experimental results show that DCNN-RRDA achieved 98.54% accuracy, 97.46% precision, 96.97% recall, and 95.88% F1-score, consistently outperforming baseline models under challenging conditions such as noise, occlusion, and illumination changes. The proposed hybrid approach demonstrates that combining deep hierarchical learning with evolutionary optimization can significantly improve recognition reliability, generalization, and robustness. These results suggest its practical applicability in real-world intelligent security systems, while highlighting the potential for further enhancements in real-time deployment and lightweight implementations for resource-constrained environments.

Deep learning (also known as deep structured learning, hierarchical learning, or deep machine learning) is a branch of machine learning based on a set of algorithms that attempt to model high-level abstractions in data by using a deep graph with multiple processing layers, composed of multiple linear and nonlinear transformations. Deep learning has been characterized as a class of machine learning algorithms with the following characteristics [257]: • They use a cascade of many layers of nonlinear processing units for feature extraction and transformation. Each successive layer uses the output from the previous layer as input. The algorithms may be supervised or unsupervised and applications include pattern analysis (unsupervised) and classification (supervised). • They are based on the (unsupervised) learning of multiple levels of features or representations of the data. Higher-level features are derived from lower-level features to form a hierarchical representation. • They are part of the broader machine learning field of learning representations of data. • They learn multiple levels of representations that correspond to different levels of abstraction; the levels form a hierarchy of concepts. These definitions have in common: multiple layers of nonlinear processing units and the supervised or unsupervised learning of feature representations in each layer, with the layers forming a hierarchy from low-level to high-level features. Various deep learning architectures such as deep neural networks, convolutional deep neural networks, deep belief networks (DBN), and recurrent neural networks have been applied to fields like computer vision, automatic speech recognition, natural language processing, audio recognition, and bioinformatics where they have been shown to produce state-of-the-art results on various tasks. In this chapter we start in Section 7.1 with an introduction, giving a brief history of this field, the relevant literature, and its applications. Then we study some basic concepts of deep learning such as convolutional neural networks, recurrent neural networks, backpropagation algorithm, restricted Boltzmann machines, and deep learning networks in Section 7.2. Then we illustrate three examples for Apache Spark implementation for mobile big data (MBD), user moving pattern extraction, and combination with nonparametric Bayesian learning, respectively, in Sections 7.3 through 7.5. Finally, we have summary in Section 7.6.

- Crowd density prediction is a critical task in modern public safety, urban planning, and large-scale event management. Accurately estimating the number of people and their spatial distribution enables authorities to prevent dangerous overcrowding and optimize resource deployment. This paper presents a hybrid crowd density prediction framework combining Convolutional Neural Networks (CNN) with Histogram of Oriented Gradients (HOG) feature descriptors. The CNN branch employs a modified VGG-16 backbone with dilated convolutions to capture high-level spatial representations, while the HOG branch extracts complementary gradient-based structural information from local image regions. Feature embeddings from both branches are fused through fully connected layers with dropout regularization to generate spatial density maps. The system is trained and validated on the ShanghaiTech Part A and Part B benchmark datasets. The proposed CNN+HOG hybrid achieves a Mean Absolute Error (MAE) of 75.4 and Root Mean Square Error (RMSE) of 112.3 on Part A, and MAE of 11.8, RMSE of 18.5 on Part B - outperforming pure CNN and HOG baselines. Training convergence analysis and error distribution studies confirm the robustness and generalization of the proposed approach across varying crowd densities and scene complexities. KeyWords: Crowd density prediction, convolutional neural network (CNN), histogram of oriented gradients (HOG), deep learning, density map estimation, ShanghaiTech dataset, feature fusion.

Objectives/Aim: The present paper deals with building a smart assistant with the aim of assisting the visually impaired people in mobility with confidence by realizing the nearby obstacles and also implement image processing techniques to recognize people. Methods: We use Raspberry Pi 4 which has increased computing performance and is interfaced to the picamera, GPS and GSM modules. Arduino pro mini with buzzer, ultrasonic sensors and vibration motors are used for obstacle detection. Certain libraries important for image processing are used such as OpenCV, Dlib, face detection (Haar cascades, HOG + Linear SVM, or CNNs), espeak for converting text to speech. Programming was implemented through Python and Arduino compiler. Findings: Analysis was carried out using the proposed system for the blind and visually impaired people who could move around comfortably with confidence and were also able to detect objects and recognize people. Hence the proposed system removes the use for the cumbersome white cane in exchange for small and handy modules which can be used in the form of wearables mounted on shoulders, hands and legs to detect the obstacles from multiple directions and provide for a comfortable wear. It was determined that the Raspberry pi 4 was incapable of running CNN detection for which ideally computer GPU was required, so HOG detection method was used instead. Application/Improvements: This project can be implemented mainly in the commercial field of helping visually impaired people with poor eyesight or being completely blind. Industrial applications can be devised and enhanced like robots and machineries along with Security, Identifying and Tracking. Keywords: Face Recognition; Obstacle Detection; Open Source Computer Vision (OpenCV); Deep Learning; Histogram of Oriented Gradients (HOG); Support Vector Machine (SVM); Convolutional Neural Network (CNN)

In this paper we evaluate the combination of hand-crafted and deep learning-based features for neonatal pain assessment. To this end we consider two hand-crafted descriptors, i.e. Local Binary Patterns (LBP) and Histogram of Oriented Gradients (HOG), and features extracted from two pre-trained Convolutional Neural Networks (CNNs). Experimental results on the publicly available Infant Classification Of Pain Expressions (COPE) database show competitive results compared to previous methods.