This study addresses the critical task of automatically identifying oceanic eddies, essential features for marine energy and chemical distribution, using sea surface temperature data from the Atlantic Ocean. It introduces the Deep Eddy Network, a sophisticated deep-learning framework based on an encoder-decoder architecture. The network performs pixel-wise classification, generating an output map where each pixel is labeled as '0' (non-eddy), '1' (anticyclonic eddy), or '2' (cyclonic eddy). Key innovations include a dedicated morphological module that injects shapebased information into the input data. The architecture is designed for high efficiency, employing advanced techniques in its core components. The encoder block utilizes dilated convolutions combined with activation functions, batch normalization, and an attention mechanism. Similarly, the decoder block integrates activation functions with 2D transpose convolution, batch normalization, and attention. Developed using Python and Keras, the final model demonstrates a superior balance between computational performance and segmentation accuracy. This makes the proposed Deep Eddy Network a practical and powerful tool, particularly suitable for deployment in real-time oceanographic monitoring and analysis applications, advancing our ability to understand these dynamic oceanic phenomena.

The rapid growth of video data has resulted in an increasing need for surveillance and violence detection systems. Although such events occur less frequently than normal activities, developing automated video surveillance systems for violence detection has become essential to minimize labor and time waste. Detecting violent activity in videos is a challenging task due to the variability and diversity of violent behavior, which can involve a wide range of actions, motions, and interactions between people and objects. Currently, researchers employ deep learning models to detect violent behaviors. In fact, a large number of deep learning approaches are based on extracting spatio-temporal information from a video by exploiting a 3D Convolutional Neural Network (CNN). Despite their success, these techniques require a lot more parameters than 2D CNNs and have high computational complexity. Therefore, we focus on exploiting a 2D CNN to encode spatio-temporal information. Actually, statistical features of the optical flow changes are used to give this ability to a 2D CNN. These features are designed to make attention to regions of a video clip with much more motion. Accordingly, the optical flow of an input video is calculated. To determine meaningful changes in the optical flow, the optical flow magnitude of a current frame is compared with its predecessor. After that, statistical features of these changes are extracted to summarize a video clip to a 2D template, which feeds a 2D CNN. Experimental results on four benchmark datasets observe that the suggested strategy outperforms baseline ones. In particular, we make a better estimation of the spatio-temporal features in a video by shortening a video clip into a 2D template.

Detecting violence in videos is an important task for ensuring public safety and security. It can be difficult, as it involves analyzing videos from multiple surveillance cameras at any time and place. The system should detect violent activities accurately and quickly, alerting security personnel in real time. To accomplish this, we propose a hybrid approach that combines the strengths of 2D convolutional neural networks (2D CNNs) and 3D CNNs for violence detection. Our approach also incorporates feature fusion, spatial and temporal attention mechanisms to improve the model’s accuracy. Specifically, we first use a 3D convolution layer to capture temporal features by considering the sequence of frames over time. Then, we use a 2D depth-wise convolution layer to extract spatial features from each video frame and create spatial attention. The spatial attention module is called depth-wise spatial attention (DSA). We also exploit a simple channel attention module to focus on temporal features. While the spatial and temporal attention modules are applied to the input videos, we fuse these extracted features using a proposed feature fusion method named max-pooling fusion module (MFM). Since the proposed framework relies on deep and complex modules, it requires the computational power provided by high-performance computing (HPC) systems. To evaluate our approach, it is applied to four benchmark datasets. Our experiments demonstrate that the proposed hybrid approach achieves superior performance compared to several state-of-the-art methods, making it an effective solution for violence detection in videos.

Oceanic eddies are a widespread and important occurrence that plays a vital role in the movement of chemicals and energy within the marine ecosystem. Hence, the astute and precise recognition of these swirling currents may greatly contribute to the progress of our comprehension of oceanography. Due to the continuous breakthroughs in state-of-the-art deep learning technology, the population is witnessing a progressive improvement in the methods used to identify and understand these aquatic characteristics. This study employs sea surface temperature data acquired from the Copernicus Marine and Environment Monitoring Service (CMEMS) in the Atlantic Ocean. The objective is to present EddyNet, a cutting-edge deep-learning framework specifically developed for the automatic identification and categorization of ocean eddies. EddyNet incorporates a pixel-wise classification layer into its neural encoder-decoder architecture. The resulting output is a map that maintains the same dimensions as the input, but each individual pixel is assigned a label indicating its classification as either "0" for noneddy regions, "1" for anticyclonic eddies, or "2" for cyclonic eddies. We propose a new image segmentation method based on the U-net architecture with different convolutional neural network backbones such as VGG16, VGG19, DenseNet121, and MobileNetV2. Our models are built and trained using Python and the Keras library with the Adam optimizer for improved convergence. Our approach uses sparse categorical cross-entropy as the loss function, simplifying the label encoding process for multiclass classification with sparse labels. Initial results show that this method achieves a good balance between computational efficiency and segmentation accuracy, making it suitable for real-time applications.

One of the most challenging problems in machine learning is the classification of imbalanced data. This problem arises when the samples of data are distributed unevenly among the classes, such that compared to one class (the majority or negative class), the other class (the minority or positive class) has far fewer samples. The classical classifiers are inappropriate to classify data sets of this nature. To address these classifiers’ shortcoming in class imbalance situations, we present a novel clustering-based over-sampling technique in this paper. First, the $k$ -means clustering algorithm is used to cluster the minority class samples. Then, sparse clusters including fewer samples are chosen. Finally, we use the nearest neighbor of each cluster center to create synthetic samples for the minority class. Also, to select clusters based on probabilities, we apply the roulette wheel selection operator during over-sampling. The C4.5 decision tree classifier is utilized in our experiments, and AUC and $F$ -measure metrics are considered to evaluate methods. According to the results, our method outperforms six other methods over fifteen imbalanced data sets.

Surveillance systems are essential to address the risk of criminal activities. It is crucial to detect such activities in real time to prompt an immediate police response. Currently, researchers are developing, training, and testing deep-learning models to detect these behaviors. In a violence detection system, capturing both spatial and temporal information is essential for accurately classifying violent activities. 3D Convolutional Neural Networks (CNNs) are particularly effective and common methods for extracting spatiotemporal features because they can model both spatial and temporal features in video sequences. However, 3D CNNs are known to be computationally intensive and require more training data compared to 2D CNNs. To address this problem, we propose a hybrid model by a fusion of 3D CNNs, a 2D depth-wise CNN, and a 2D CNN to classify violent activities. Employing a hybrid CNN not only reduces the complexity of the model, but it can also exhibit robust performance in a limited quantity of training samples. We exploit a 2D depth-wise CNN for spatial attention as it applies a single filter to each input channel individually without mixing information across channels. Therefore, we use it in the proposed model for spatial attention. Finally, the proposed approach is evaluated on four benchmark datasets: Hockey Fight, Surveillance Fight, Violent Flows, and Action Movies, achieving accuracy rates of 99.3%, 98.46%, 99.92%, and 98.2% respectively on each dataset.

Violence detection is a difficult task because it involves analyzing video clips from multiple security cameras, which are located in various places and operate continuously. When violent crimes occur, a system should be able to reliably detect them in real-time and immediately alert a surveillance team. Currently, researchers employ deep learning models to detect violent behavior. Notably, a large number of deep learning approaches are based on extracting spatio-temporal information from a video by exploiting either 3D Convolutional Neural Networks (CNNs) or multi-stream networks. Despite their success, these techniques require a lot of parameters than 2D CNNs and have high computational complexity. Therefore, we present a simple spatio-temporal attention mechanism combined with a 2D CNN for an effective violence detection system. We propose a Squeeze Temporal Attention block that allows a 2D CNN to learn spatiotemporal features in videos. This effective block uses squeeze and temporal attention modules to summarize a video stream into three channels. In addition, we introduce spatial attention and feature fusion modules to improve the performance of the proposed system. The spatial attention module, Entropy Spatial Module, utilizes an entropy filter and frame differences to focus on spatial regions of the video with more movement. The fusion module parallelizes two dense layers with a 2D CNN to effectively enhance the classifier's performance. As a result, our proposed model achieves improved performance results in terms of accuracy when compared to Long Short-Term Memory, multi-stream networks, and current 3D CNNs.

Recent advancements in remote sensing technology have significantly expanded the exploration of natural resources and enabled the detection of materials in inaccessible areas. Hyperspectral images (HSIs) are a valuable data source due to their distinctive properties in various applications. However, several problems, including noise, band correlation, ineffectively extracted features, and most notably, a lack of sufficient labeled samples, reduce the accuracy of HSI classification. To improve the performance of such a system, we propose an effective method with the capability of paying attention to spectral and spatial features. The raw HSI data are first preprocessed using a principal component analysis (PCA) operation because of the redundancy and correlation between HSI bands. Then, the entropy base informative module is designed to add entropy information to the selected spectral features by PCA. We also use spectral and spatial attention modules in the proposed model. Moreover, a hybrid neural network that uses both 3-D convolutional neural networks (CNNs) and 2-D CNNs with skip connections is exploited to reduce the complexity of the network compared to 3-D CNNs. The spatial attention module called depthwise spatial attention block can inherently highlight spatial information. The spectral attention module named reshape softmax attention can capture useful spectral regions of feature maps. Meticulous HSI classification tests are conducted over the University of Pavia, Indian Pines, Salinas, and Houston 2013 to evaluate the effectiveness of our approach. Our experiments show higher accuracy compared to other deep learning methods.

With the rapid development of detecting violent behaviors in surveillance cameras, requests on systems that automatically recognize violent events are expanded. Nowadays, violence detection has become an active research field in image processing and machine learning. The relevant works in such a field are classified into hand-crafted and deep learning methods. Despite the effectiveness of hand-crafted ones, their computational cost may be suppressive for practical applications. Additionally, deep learning techniques usually exploit 3D Convolutional Networks (3D ConvNets) to do this task. To improve the accuracy of these networks, meaningful regions and temporal changes in videos should be considered. Consequently, the performance of a 3D ConvNet can be reinforced by selecting significant temporal information and noticing to special regions in two spatial dimensions. In this work, we propose a novel 3D ConvNet along with a technique for extracting interest frames. The Structural Similarity Index Measure (SSIM) is exploited to extract interest frames as significant temporal information. Indeed, the SSIM uses the statistical features of two consecutive frames for this reason. In this way, sixteen video frames with the smallest SSIM are considered as dominant motion frames, which are then sent to a 3D CNN for classification. Moreover, a spatial attention module is exploited to make attention on the specific regions. Furthermore, three benchmark datasets are employed to evaluate the performance of the proposed method. The results show that in terms of accuracy, our scheme outperforms existing approaches.

Violence detection is one of the substantial and challenging topics in intelligent video surveillance systems. As there is a growing demand on video surveillance systems with the capability of automatic violence detection, we focus on existing violence detection methods to improve them. In this paper, we introduce a new feature descriptor named Histogram of Optical flow Magnitude and Orientation (HOMO). First, the proposed method converts input frames to the grayscale format. Next, it computes the optical flow between two consequence frames. Then, the optical flow magnitude and orientation of each pixel in each frame are compared separately with its predecessor frame to obtain meaningful changes of magnitude and orientation. Subsequently, different threshold values are applied to the magnitude and orientation changes for obtaining six binary indicators. Finally, these binary indicators are analyzed to get the HOMO descriptor which is used to train a SVM classifier. The system has been implemented using MATLAB. To evaluate the proposed method, two benchmark datasets have been used. The comparison of HOMO and other descriptors on benchmark datasets demonstrates satisfactory performance. (C) 2019 Elsevier Ltd. All rights reserved.

This paper presents license plate (LP) detection and recognition of Iranian LP digits. The proposed method can be divided into four major steps which are preprocessing, digit segmentation, feature extraction and finally classification using naive Bayesian (NB) classifier. In the preprocessing step, the obtained vehicle images are converted to the binary format based on a proposed threshold value. In the digit segmentation step, the LP digits are extracted from the image based on connected component labeling and some extracted characteristics of LP digits. In the feature extraction step, the self-organizing feature maps (SOFM) is used. In the classification step, the digits are recognized by a NB classifier which its performance is compared with a K-NN classifier. Various images in different conditions were used to test the proposed algorithm and experimental results demonstrated its robustness.

This paper presents the automatic recognition of two Iranian accidental music notes from music sheets and their distinct classification from three western accidental notes. The proposed method is designed for recognizing handwritten documents it consists of four major steps, preprocessing, note extraction, feature extraction and finally classification using Artificial Neural Networks (ANNs). In the preprocessing step, the noise is removed and the thickness of staff lines is extracted then the morphological operation is used to remove staff lines. In the note extraction step, the notes are extracted from the scanned music sheets by connected component labeling. In the third step, the zoning feature and Hough transform are used to extract features. Finally at the last step, the notes are classified by the ANNs. The developed procedure was applied to the different music sheets scanned by a scanner with 300 dpi resolution and the recognition rate of accidental notes was 100%.

Robot path planning is an important part of the development of autonomous systems. Numerous strategies have been proposed in the literature regarding mobile robots but trajectory planning for manipulators is considerably more difficult since the entire structure can move and therefore produce collisions with surrounding obstacles. This paper presents an original solution and analytical comparison to path planning for manipulator arms. Path planning is executed in two parts: first, a global path is found to guide the end-effector in the environment using artificial potential fields and multi-resolution occupancy grids, then, a local path is determined for the entire robot structure by considering the kinematics of the robot as well as the repulsive forces of nearby obstacles in a fuzzy logic controller. Results are shown from a simulator that has been built for this purpose. The contribution of this research is to develop a robust solution for path planning with collision avoidance: one that can be used for various manipulator arms and environment configurations.

Cognitive method is used in this research to create portfilo of movement robot manipulator. Gradient descent (GD) fuzzy controller was used and robot’s postures and trajectory were expected in MATLAB/SIMULINK environment. Fuzzy logic controller (FLC) is an influential nonlinear controller to certain and uncertain systems which it is based on artificial intelligent and computes the required torques using the nonlinear feedback control law. Practically a large amount of systems have uncertainties accordingly this method has a challenge. Linear proportional-integral-derivative (PID) controller is a significant combination nonlinear stable-robust controller under condition of partly uncertain dynamic parameters of system. This technique is used to control of highly nonlinear systems especially in nonlinear time varient nonlinear dynamic system. To compensate for the dependence on model-free parameters, PID methodology is used. The nonlinear model dynamic formulation problem in uncertain system can be solved by using PID theorem. Proportional-DerivativeIntegral (PID) theory is used to estimate the system dynamic.

Optical music recognition (OMR) is one of the applications of image processing, pattern recognition and music theory. Skew detection is an essential step in OMR system in order to enhance the performance of OMR system. This paper presents two new algorithms for skew detection in music score image. It includes two skew detection algorithms based on moments and skeleton. The proposed technique consists of three major steps, image smoothing, musical symbols removal and finally skew detection algorithms. In the image smoothing step, the enhancement techniques are used to provide a more appropriate image for the subsequent steps. In the musical symbol removal step, musical symbols are removed by morphological operation in order to extract staff lines. In the last step, skew detection algorithms, the proposed method are applied to the staff lines image. The experimental results show that the proposed technique is robust for skew detection.

This paper proposes a new algorithm for skew detection in music score images. A common method for finding location of staff lines is horizontal projection that is very sensitive to skew. The proposed method consists of four stages, preprocessing, staff line detection, region of interest (ROI) selection and skew detection. After preprocessing, some parameters are extracted for staff line detection and ROI selection. By using ROI, execution time of skew detection is reduced. The proposed method has been applied to several music sheets and skew is detected with a high precision.