Illumination Variations Research Articles

Framework for abnormal event detection and tracking based on effective sparse factorization strategy

The idea of tracking video objects has evolved to facilitate the area of surveillance systems. However, most current research efforts lie in speedy abnormal event detection and tracking of objects of interest tracking. However, the primary challenge is dealing with complex video structures' inherent redundancy. The existing research models for video tracking are more inclined towards improving accuracy. In contrast, the consideration of a more significant proportion of mobile object dynamics, e.g. abnormal events, in motion over the crowded video frame sequence is mainly overlooked, which is essential to study a specific movement pattern of the object of interest appearing in the frame sequence concerning the cost of computation factors. The study thereby introduces a unique strategy of speedy abnormal event detection and tracking, which facilitates video tracking to assess a specific pattern of object of interest movement over complex and crowded video scenes, considering a unique learning-based approach. The extensive simulation outcome further shows that the proposed tracking model accomplishes better tracking accuracy yet retains an optimized computation cost compared to the baseline studies. The computation of video tracking also accomplishes higher detection rates even in the challenging constraints of partial/complete occlusion, illumination variation and background clutter.

Selective weighted least square and piecewise bilinear transformation for accurate satellite DSM generation

One of the main products of multi-view stereo (MVS) high-resolution satellite (HRS) images in photogrammetry and remote sensing is digital surface model (DSM). Producing DSMs from MVS HRS images still faces serious challenges due to various reasons such as complexity of imaging geometry and exterior orientation model in HRS, as well as large dimensions and various geometric and illumination variations. The main motivation for conducting this research is to provide a novel and efficient method that enhances the accuracy and completeness of extracting DSM from HRS images compared to existing recent methods. The proposed method called Sat-DSM, consists of five main stages. Initially, a very dense set of tie-points is extracted from the images using a tile-based matching method, phase congruency-based feature detectors and descriptors, and a local geometric consistency correspondence method. Then, the process of Rational Polynomial Coefficients (RPC) block adjustment is performed to compensate the RPC bias errors. After that, a dense matching process is performed to generate 3D point clouds for each pair of input HRS images using a new geometric transformation called PWB (pricewise bilinear) and an accurate area-based matching method called SWLSM (selective weighted least square matching). The key innovations of this research include the introduction of SWLSM and PWB methods for an accurate dense matching process. The PWB is a novel and simple piecewise geometric transformation model based on superpixel over-segmentation that has been proposed for accurate registration of each pair of HRS images. The SWLSM matching method is based on phase congruency measure and a selection strategy to improve the well-known LSM (least square matching) performance. After dense matching process, the final stage is spatial intersection to generate 3D point clouds, followed by elevation interpolation to produce DSM. To evaluate the Sat-DSM method, 12 sets of MVS-HRS data from IRS-P5, ZY3-1, ZY3-2, and Worldview-3 sensors were selected from areas with different landscapes such as urban, mountainous, and agricultural areas The results indicate the superiority of the proposed Sat-DSM method over four other methods CATALYST, SGM (Semi-global matching), SS-DSM (structural similarity based DSM extraction), and Sat-MVSF in terms of completeness, RMSE, and MEE. The demo code is available at https://www.researchgate.net/publication/377721674_SatDSM.

Unveiling Illumination Variations During a Lunar Eclipse: Multi-Wavelength Spaceborne Observations of the January 21, 2019 Event

Space-based observations of the total lunar eclipse on 21 January 2019 were conducted using the geostationary Earth-orbiting satellite Gaofen-4 (GF-4). This study represents a pioneering effort to address the observational gap in full-disk lunar eclipse photometry from space. With its high resolution and ability to capture the entire lunar disk, GF-4 enabled both quantitative and qualitative analyses of the variations in lunar brightness, as well as spectra and color changes, across two spatial dimensions, from the whole lunar disk to resolved regions. Our results indicate that before the totality phase of the lunar eclipse, the irradiance of the Moon diminishes to below approximately 0.19% of that of the uneclipsed Moon. Additionally, we observed an increase in lunar brightness at the initial entry into the penumbra. This phenomenon is attributed to the opposition effect, providing scientific evidence for this unexpected behavior. To investigate detailed spectral variations, specific calibration sites, including the Chang’E-3 landing site, MS-2 in Mare Serenitatis, and the Apollo 16 highlands, were analyzed. Notably, the red-to-blue ratio dropped below 1 near the umbra, contradicting the common perception that the Moon appears red during lunar eclipses. The red/blue ratio images reveal that as the Moon enters Earth’s umbra, it does not simply turn red; instead, a blue-banded ring appears at the boundary due to ozone absorption and the lunar surface composition. These findings significantly enhance our understanding of atmospheric effects on lunar eclipses and provide crucial reference information for the future modeling of lunar eclipse radiation, promoting the integration of remote sensing science with astronomy.

Novel Approach in Vegetation Detection Using Multi-Scale Convolutional Neural Network

Vegetation segmentation plays a crucial role in accurately monitoring and analyzing vegetation cover, growth patterns, and changes over time, which in turn contributes to environmental studies, land management, and assessing the impact of climate change. This study explores the potential of a multi-scale convolutional neural network (MSCNN) design for object classification, specifically focusing on vegetation detection. The MSCNN is designed to integrate multi-scale feature extraction and attention mechanisms, enabling the model to capture both fine and coarse vegetation patterns effectively. Moreover, the MSCNN architecture integrates multiple convolutional layers with varying kernel sizes (3 × 3, 5 × 5, and 7 × 7), enabling the model to extract features at different scales, which is vital for identifying diverse vegetation patterns across various landscapes. Vegetation detection is demonstrated using three diverse datasets: the CamVid dataset, the FloodNet dataset, and the multispectral RIT-18 dataset. These datasets present a range of challenges, including variations in illumination, the presence of shadows, occlusion, scale differences, and cluttered backgrounds, which are common in real-world scenarios. The MSCNN architecture allows for the integration of information from multiple scales, facilitating the detection of diverse vegetation types under varying conditions. The performance of the proposed MSCNN method is rigorously evaluated and compared against state-of-the-art techniques in the field. Comprehensive experiments showcase the effectiveness of the approach, highlighting its robustness in accurately segmenting and classifying vegetation even in complex environments. The results indicate that the MSCNN design significantly outperforms traditional methods, achieving a remarkable global accuracy and boundary F1 score (BF score) of up to 98%. This superior performance underscores the MSCNN’s capability to enhance vegetation detection in imagery, making it a promising tool for applications in environmental monitoring and land use management.

An Efficient Dense Reconstruction Algorithm from LiDAR and Monocular Camera

Dense reconstruction have been studied for decades in the fields of computer vision and robotics, in which LiDAR and camera are widely used. However, vision-based methods are sensitive to illumination variation and lack direct depth, and LiDAR-based methods are limited by sparse LiDAR measurement and lacking color and texture information. In this paper, we propose a novel 3D reconstruction algorithm based on LiDAR and a monocular camera, which realizes dense reconstruction. In the algorithm, a LiDAR odometry is used to get accurate poses and poses calculated by the odometry module are used in the calculation of depth maps and fusion of depth maps, and then mesh and texture mapping are implemented. In addition, a semantic segmentation network and a depth completion network are used to obtain dense and accurate depth maps. The concept of symmetry is utilized to generate 3D models of objects or scenes; that is, the reconstruction and camera imaging of these objects or scenes are symmetrical. Experimental results on public dataset show that the proposed algorithm achieves higher accuracy, efficiency and completeness than existing methods.

PGSR: Planar-based Gaussian Splatting for Efficient and High-Fidelity Surface Reconstruction.

Recently, 3D Gaussian Splatting (3DGS) has attracted widespread attention due to its high-quality rendering, and ultra-fast training and rendering speed. However, due to the unstructured and irregular nature of Gaussian point clouds, it is difficult to guarantee geometric reconstruction accuracy and multi-view consistency simply by relying on image reconstruction loss. Although many studies on surface reconstruction based on 3DGS have emerged recently, the quality of their meshes is generally unsatisfactory. To address this problem, we propose a fast planar-based Gaussian splatting reconstruction representation (PGSR) to achieve high-fidelity surface reconstruction while ensuring high-quality rendering. Specifically, we first introduce an unbiased depth rendering method, which directly renders the distance from the camera origin to the Gaussian plane and the corresponding normal map based on the Gaussian distribution of the point cloud, and divides the two to obtain the unbiased depth. We then introduce single-view geometric, multi-view photometric, and geometric regularization to preserve global geometric accuracy. We also propose a camera exposure compensation model to cope with scenes with large illumination variations. Experiments on indoor and outdoor scenes show that the proposed method achieves fast training and rendering while maintaining high-fidelity rendering and geometric reconstruction, outperforming 3DGS-based and NeRF-based methods. Our code will be made publicly available, and more information can be found on our project page (https://zju3dv.github.io/pgsr/).

EFFECTIVE FEATURE EXTRACTION METHOD FOR UNCONSTRAINED ENVIRONMENT: LOCAL BINARY PATTERN OR LOCAL TERNARY PATTERN

In this study, a range of algorithms addressing the challenges posed by noise and illumination were investigated. Two algorithms, namely LTP and LBP, were selected for comparison due to their demonstrated effectiveness. The process becomes time-consuming due to training samples, mainly when dealing with images featuring higher levels of noise and illumination variations, necessitating efficient algorithms for effective recognition. To compare two effective feature extraction methods viz local binary pattern (LBP) and local ternary pattern (LTP) for an unconstraint environment. The impact of noise and illumination factors is particularly pronounced in the iris datasets of non-cooperative subjects, which serve as the input images for this analysis. These algorithms were applied to diverse datasets with distinctive illumination properties to facilitate feature extraction. The results indicated that the LTP exhibited efficiency in comparison, suggesting its efficacy in handling datasets with varying illumination characteristics. A comparative analysis between LBP and LTP was conducted on two distinct datasets, namely UBIRIS and CASIA. The investigation into the sensitivity of LTP revealed heightened sensitivity during the performance analysis test, with consistent accuracy observed at 50 samples and a scale of 0.3. In the case of the CASIA iris dataset, the recital of LTP and LBP exhibited nearly identical accuracy levels, converging after 70 samples for non-cooperative iris datasets compared to the LBP.

Enhancing the MCL-SLAM algorithm to overcome the issue of illumination variation, non-static environment and kidnapping to present the NIK-SLAM

AbstractSimultaneous Localization and Mapping (SLAM) is a major challenge that has been comprehensively investigated by scholars. SLAM’s ability to support autonomous navigation helps the robot to move from one place to another without being controlled by humans. This achievement has attracted the attention of scholars towards focusing on this area. Over the decades, countless procedures have been developed with extraordinary accomplishments. However, some problems can degrade the efficiency of SLAM (Simultaneous Localization and Mapping) technique, and examples of such problems include environmental noises (light intensity and shadow), non-static environment, and kidnap robots. These problems create inconsistency in measurement that can lead to erroneous navigation for an unmanned vehicle. In minimizing the effect of these problems, a novel SLAM technique based on several re-modifications of MCL is presented in this paper. This innovative technique is known as NIK-SLAM. The NIK-SLAM is equipped with filters alongside some modifications of the original Monte-Carlo algorithm (MCL) to maximize its performance. The filters were employed to overcome the effect of shadow and light intensity while the re-modification of MCL is used to overcome the limitation of kidnapping and non-static objects. A publicly available dataset (TUM-RGBD) and a private dataset were used for evaluation. Matlab was used for simulation while assessments were based on quantitative and qualitative validation. The results demonstrated a better performance of the NIK-SLAM, attaining a lower trajectory error for most experiments when compared to the original MCL and other SLAM algorithms found in the literature, but at the expense of delayed processing time. Furthermore, this accomplishment supports pathway planning, and exploration into unstable environments while it decreases accident rates/human fatalities, and in the future study, the issue of delayed processing time will be addressed.

An improved YOLOv8n-IRP model for natural rubber tree tapping surface detection and tapping key point positioning.

Aiming at the problem that lightweight algorithm models are difficult to accurately detect and locate tapping surfaces and tapping key points in complex rubber forest environments, this paper proposes an improved YOLOv8n-IRP model based on the YOLOv8n-Pose. First, the receptive field attention mechanism is introduced into the backbone network to enhance the feature extraction ability of the tapping surface. Secondly, the AFPN structure is used to reduce the loss and degradation of the low-level and high-level feature information. Finally, this paper designs a dual-branch key point detection head to improve the screening ability of key point features in the tapping surface. In the detection performance comparison experiment, the YOLOv8n-IRP improves the D_mAP50 and P_mAP50 by 1.4% and 2.3%, respectively, over the original model while achieving an average detection success rate of 87% in the variable illumination test, which demonstrates enhanced robustness. In the positioning performance comparison experiment, the YOLOv8n-IRP achieves an overall better localization performance than YOLOv8n-Pose and YOLOv5n-Pose, realizing an average Euclidean distance error of less than 40 pixels. In summary, YOLOv8n-IRP shows excellent detection and positioning performance, which not only provides a new method for the key point localization of the rubber-tapping robot but also provides technical support for the unmanned rubber-tapping operation of the intelligent rubber-tapping robot.

Research on the Registration of Aerial Images of Cyclobalanopsis Natural Forest Based on Optimized Fast Sample Consensus Point Matching with SIFT Features

The effective management and conservation of forest resources hinge on accurate monitoring. Nonetheless, individual remote-sensing images captured by low-altitude unmanned aerial vehicles (UAVs) fail to encapsulate the entirety of a forest’s characteristics. The application of image-stitching technology to high-resolution drone imagery facilitates a prompt evaluation of forest resources, encompassing quantity, quality, and spatial distribution. This study introduces an improved SIFT algorithm designed to tackle the challenges of low matching rates and prolonged registration times encountered with forest images characterized by dense textures. By implementing the SIFT-OCT (SIFT omitting the initial scale space) approach, the algorithm bypasses the initial scale space, thereby reducing the number of ineffective feature points and augmenting processing efficiency. To bolster the SIFT algorithm’s resilience against rotation and illumination variations, and to furnish supplementary information for registration even when fewer valid feature points are available, a gradient location and orientation histogram (GLOH) descriptor is integrated. For feature matching, the more computationally efficient Manhattan distance is utilized to filter feature points, which further optimizes efficiency. The fast sample consensus (FSC) algorithm is then applied to remove mismatched point pairs, thus refining registration accuracy. This research also investigates the influence of vegetation coverage and image overlap rates on the algorithm’s efficacy, using five sets of Cyclobalanopsis natural forest images. Experimental outcomes reveal that the proposed method significantly reduces registration time by an average of 3.66 times compared to that of SIFT, 1.71 times compared to that of SIFT-OCT, 5.67 times compared to that of PSO-SIFT, and 3.42 times compared to that of KAZE, demonstrating its superior performance.

AVENUE: A novel deep fake detection method based on temporal convolution network & rPPG information

In Deep Learning (DL), an adversary creates Deepfakes by manipulating facial features to fool someone. The Deepfakes pose a security threat to anyone's privacy and a primary concern for our society. It can be detected by utilizing the texture and physiological properties of the face, like eye and lip movements; however, such methods are incompetent when Deepfakes are created using recent generative adversarial networks (GAN). Alternatively, remote Photoplethysmography (rPPG) information can be used for Deepfake detection because GANs neglect human physiological information for Deepfake generation. Such detection can be inaccurate when rPPG signals are affected by the noises induced by facial deformation and illumination variations. Furthermore, the exiting Deepfake detections are usually performed using sequential models, and such models fail to process the long sequence of temporal information. These issues are mitigated by our proposed method AVENUE , that is, \(A\) no \(V\) el d \(E\) epfake detectio \(N\) method based on temporal convol \(U\) tion n \(E\) twork & rPPG information. For mitigating the noise issues in the rPPG signals, the proposed method detects and employs relatively stable clips of the input video for Deepfake detection. The stable clips are those clips that are least affected by facial deformations. Also, we use a modified Temporal convolution network to model the long sequence of Deepfake information rather than the sequential architectures. We performed the experimental result on publically available datasets of Deepfake videos. It demonstrates that our proposed method performs better than the existing rPPG-based Deepfake detection methods.

Regional response to light illuminance across the human hypothalamus.

Light exerts multiple non-image-forming biological effects on physiology including the stimulation of alertness and cognition. However, the subcortical circuitry underlying the stimulating impact of light is not established in humans. We used 7 Tesla functional magnetic resonance imaging to assess the impact of variations in light illuminance on the regional activity of the hypothalamus while healthy young adults (N=26; 16 women; 24.3±2.9 y) were completing two auditory cognitive tasks. We find that, during both the executive and emotional tasks, higher illuminance triggered an activity increase over the posterior part of the hypothalamus, which includes part of the tuberomamillary nucleus and the posterior part of the lateral hypothalamus. In contrast, increasing illuminance evoked a decrease in activity over the anterior and ventral parts of the hypothalamus, encompassing notably the suprachiasmatic nucleus and another part of the tuberomammillary nucleus. Critically, the performance of the executive task was improved under higher illuminance and was negatively correlated with the activity of the posterior hypothalamus area. These findings reveal the distinct local dynamics of different hypothalamus regions that underlie the impact of light on cognition.

Regional response to light illuminance across the human hypothalamus

Light exerts multiple non-image-forming biological effects on physiology including the stimulation of alertness and cognition. However, the subcortical circuitry underlying the stimulating impact of light is not established in humans. We used 7 Tesla functional magnetic resonance imaging to assess the impact of variations in light illuminance on the regional activity of the hypothalamus while healthy young adults (N=26; 16 women; 24.3±2.9 y) were completing two auditory cognitive tasks. We find that, during both the executive and emotional tasks, higher illuminance triggered an activity increase over the posterior part of the hypothalamus, which includes part of the tuberomamillary nucleus and the posterior part of the lateral hypothalamus. In contrast, increasing illuminance evoked a decrease in activity over the anterior and ventral parts of the hypothalamus, encompassing notably the suprachiasmatic nucleus and another part of the tuberomammillary nucleus. Critically, the performance of the executive task was improved under higher illuminance and was negatively correlated with the activity of the posterior hypothalamus area. These findings reveal the distinct local dynamics of different hypothalamus regions that underlie the impact of light on cognition.

HGLeNet DEEP LEARNING FOR SKIN CANCER DETECTION WITH SKIN LESION SEGMENTATION

Skin cancer is deemed to be the most dangerous type of cancer. It occurs due to the damage caused to the DNA and if left untreated can lead to death. Various methods have been devised over the last few years for skin cancer detection. However, their performance was affected by various challenges in image analysis, like color illumination variations, differences in shape, size, etc. Therefore, to tackle these issues the novel framework, deep learning (DL) technique for accurate skin cancer detection and lesion segmentation is developed. Primarily, the skin image is pre-processed by employing a weighted median filtering to eradicate the noises contained. Then, the segmentation of skin lesions is carried out by employing the efficient neural network (ENet). After that, augmentation is accomplished to avoid overfitting, and later, feature extraction is carried out. At last, skin cancer detection is effectuated with the help of hybrid GoogleNet-LeNet (HGLeNet), which is obtained by merging GoogleNet and LeNet, here layers are modified using the regression concept. Furthermore, the introduced framework is examined for its effectiveness through accuracy, sensitivity and specificity. Moreover, HGLeNet attained the highest accuracy of 0.922, sensitivity of 0.928, as well as specificity of 0.924.

A Nighttime Driving-Scene Segmentation Method Based on Light-Enhanced Network

To solve the semantic segmentation problem of night driving-scene images, which often have low brightness, low contrast, and uneven illumination, a nighttime driving-scene segmentation method based on a light-enhanced network was proposed. Firstly, we designed a light enhancement network, which comprises two parts: a color correction module and a parameter predictor. The color correction module mitigates the impact of illumination variations on the segmentation network by adjusting the color information of the image. Meanwhile, the parameter predictor accurately predicts the parameters of the image filter through the analysis of global content, including factors such as brightness, contrast, hue, and exposure level, thereby effectively enhancing the image quality. Subsequently, the output of the light enhancement network is input into the segmentation network to obtain the final segmentation prediction. Experimental results show that the proposed method achieves mean Intersection over Union (mIoU) values of 59.4% on the Dark Zurich-test dataset, outperforming other segmentation algorithms for nighttime driving-scenes.

An enhanced method for reconstruction of full SIF spectrum for near-ground measurements

Recently the applications of remotely sensed Solar-Induced chlorophyll Fluorescence (SIF) in the study of photosynthesis, stress conditions, and gross primary production have increased significantly. The full SIF spectrum spans over a spectral region of 650 ∼ 850 nm with two characteristic peaks around 685 nm and 740 nm. Over recent decades, many retrieval algorithms have been developed to estimate SIF at Top-Of-Canopy (TOC) using in-situ measurements of solar irradiance and canopy radiance spectra. Although the majority of retrieval methods retrieve SIF at a narrow spectral window, there exists a potential for retrieval of SIF in the full emission spectrum. Moreover, solar irradiance and canopy radiance spectra should ideally be measured at the same time but are usually measured sequentially with a certain time lag, raising potential errors in SIF retrieval. In this study, an enhanced retrieval algorithm of the full SIF spectrum at TOC is proposed. The proposed algorithm attempts to minimize the errors owing to time mismatch in measurements of solar irradiance and canopy radiance spectra. As an improvement to the previous algorithm (advanced Fluorescence Spectrum Reconstruction, aFSR), this proposed algorithm (aFSR-SVE) models the SIF-free contribution with principal components using the singular value decomposition technique. The optimal parameter settings in the forward model were determined for the experimental data collected by spectrometers used in the study. Firstly, the proposed algorithm was used to reconstruct full SIF spectrum for simulated data. The results were compared with known reference SIF values. After achieving satisfying results from simulated data, the proposed algorithm was compared with retrievals from established algorithms using experimental data. The results show improved SIF retrieval accuracy, without the need to simultaneously measure solar irradiance and canopy radiance spectra. The retrieval values comply with the results of previous algorithms in terms of spectral shape, diurnal trend, and temporal variations. The induced errors in SIF retrievals due to non-simultaneous measurements of solar irradiance and canopy radiance spectra were also investigated and the proposed algorithm was found to be less prone to such errors. Hence, the proposed algorithm is an improvement in reconstructing the full SIF spectrum with near-ground measurements. With the help of the proposed algorithm, field measurements using sequential systems and automated measurements of multiple targets can be performed effectively as it relaxes the requirement of concurrent measurement of solar irradiance and canopy radiance spectra. For future work, the applicability of this method can be investigated under more variable illumination conditions, like high cirrus clouds, passing clouds or persistent thin clouds.

A Comparative Model for Blurred Text Detection in Wild Scene Using Independent Component Analysis (ICA) and Enhanced Genetic Algorithm (Using a Bird Approach) with Classifiers

The advent of the study of Scene Text Detection and Recognition has exposed some significant challenges text recognition faces, such as blurred text detection. This study proposes a comparative model for detecting blurred text in wild scenes using independent component analysis (ICA) and enhanced genetic algorithm (E-GA) with support vector machine (SVM) and k-nearest neighbors (KNN) as classifiers. The proposed model aims to improve the accuracy of blurred text detection in challenging environments with complex backgrounds, noise, and illumination variations. The proposed model consists of three main stages: preprocessing, feature extraction, and classification. In the preprocessing stage, the input image is first preprocessed to remove noise and enhance edges using a median filter and a Sobel filter, respectively. Then, the blurred text regions are extracted using the Laplacian of Gaussian (LoG) filter. In the feature extraction stage, ICA is used to extract independent components from the blurred text regions. The extracted components are then fed into an E-GA-based feature selection algorithm to select the most discriminative features. The E-GA simply fine tunes the selection functionalities of the traditional GA using a bird approach. The selected features are then normalized and fed into the SVM and KNN classifiers. Experimental results on a benchmarking dataset (ICDAR 2019 LSVT) shows that the model outperforms state-of-the-art methods in terms of detection accuracy, precision, recall, and F1-score. The proposed model achieves an overall accuracy of 95.13% for SVM and 88.69% for KNN, which is significantly higher than the already existing methods which for SVM is 93%. In conclusion, the proposed model provides a promising approach for detecting blurred text in wild scenes. The combination of ICA, E-GA, and SVM/KNN classifiers enhances the robustness and accuracy of the detection system, which can be beneficial for a wide range of applications, such as text recognition, document analysis, and security systems.

Robust Face Recognition System Based on Deep Facial Feature Extraction and Machine Learning

Facial Recognition is a common challenge in research for security applications. Wide-ranging applications of this technology include biometrics, security data, access to controlled locations, smart cards, and surveillance systems. A practical approach to identifying individuals depends on factors such as partial face occlusion, illumination, age, expression, makeup, and poses. These complex face recognition variables affect most face recognition systems. In this paper, we propose a deep transfer learning for facial features based on the pose and illumination variations of celebrities’ faces in a free environment for face recognition. The new approach that combines deep learning methods for feature extraction with machine learning classification methods. Also, studies the performance of the pre-trained model (VGGFace2 with ResNet-50 model weights) for feature extraction with a Multilayer Perceptron Classifier (MLP), Decision Tree, and Bootstrap Aggregation (Bagging) to perform classification. The convolution neural network has lately achieved excellent advancement in facial recognition. The outcome indicated that VGGFace2 with MLP obtained more precision and provided the best results (Accuracy, F-measure, Recall, and Precision). The results for all metric for the proposed model is 99%. The benchmark for this model is a dataset of 105 people’s faces.

Derived Amharic alphabet sign language recognition using machine learning methods

Hearing-impaired people use sign language as a means of communication with those with no hearing disability. It is therefore difficult to communicate with hearing impaired people without the expertise of a signer or knowledge of sign language. As a result, technologies that understands sign language are required to bridge the communication gap between those that have hearing impairments and those that dont. Ethiopian Amharic alphabets sign language (EAMASL) is different from other countries sign languages because Amharic Language is spoken in Ethiopia and has a number of complex alphabets. Presently in Ethiopia, just a few studies on AMASL have been conducted. Previous works, on the other hand, only worked on basic and a few derived Amharic alphabet signs. To solve this challenge, in this paper, we propose Machine Learning techniques such as Support Vector Machine (SVM) with Convolutional Neural Network (CNN), Histogram of Oriented Gradients (HOG), and their hybrid features to recognize the remaining derived Amharic alphabet signs. Because CNN is good for rotation and translation of signs, and HOG works well for low quality data under strong illumination variation and a small quantity of training data, the two have been combined for feature extraction. CNN (Softmax) was utilized as a classifier for normalized hybrid features in addition to SVM. SVM model using CNN, HOG, normalized, and non-normalized hybrid feature vectors achieved an accuracy of 89.02%, 95.42%, 97.40%, and 93.61% using 10-fold cross validation, respectively. With the normalized hybrid features, the other classifier, CNN (sofmax), produced a 93.55% accuracy.

Recognition of vehicle license plates in highway scenes with deep fusion network and connectionist temporal classification

AbstractLicense plate recognition is crucial in Intelligent Transportation Systems (ITS) for vehicle management, traffic monitoring, and security inspection. In highway scenarios, this task faces challenges such as diversity, blurriness, occlusion, and illumination variation of license plates. This article explores Recurrent Neural Networks based on Connectionist Temporal Classification (RNN‐CTC) for license plate recognition in challenging highway conditions. Four neural network models: ResNet50, ResNeXt, InceptionV3, and SENet, all combined with RNN‐CTC are comparatively evaluated. Furthermore, a novel architecture named ResNet50 Deep Fusion Network using Connectionist Temporal Classification (ResNet50‐DFN‐CTC) is proposed. Comparative and ablation experiments are conducted using the Highway License Plate Dataset of Southeast University (HLPD‐SU). Results demonstrate the superior performance of ResNet50‐DFN‐CTC in challenging highway conditions, achieving 93.158% accuracy with a processing time of 7.91 ms, outperforming other tested models. This research contributes to advancing license plate recognition technology for real‐world highway applications under adverse conditions.