Nowadays, mobile robots are applied in a variety of fields to assist humans in doing complicated work such as disaster situations, rescue missions, security and tracking systems, military campaigns, and planetary discovery. When performing these tasks, climbing uneven terrain and detecting targets are two of the most important requirements for the mobile robot. The Rocker-Bogie design developed by the NASA Mars Exploration Rover (MER) Project has become a proven mobility application during the last decade. The stability and obstacle-climbing capability of this model are evaluated as suitable choices for mobile robots. In this paper, a prototype of a Rocker-Bogie mobile robot featuring a human tracking system is proposed. To be more detailed, the modified YOLOv8 network is designed to detect humans and corresponds to the operation of the robot in real time. The computer then computes the associated speeds of the robot according to the detection targets. The velocity signal is transmitted back to the robot, which then executes the appropriate maneuver. The experimental result demonstrates an average success rate of 91% for tracking missions on an actual mobile robot platform under different scenarios.

Detecting small objects is a significant challenge when evaluating images captured from UAVs. Although YOLO models have been successful in detecting conventional objects, but they still face challenges in detecting small objects in traffic monitoring scenarios. Applying the Slicing-Aided Hyper Inference (SAHI) framework to YOLO models can expand the pixel regions containing small objects, thereby improving the detection of objects with tiny and small sizes. However, this can also lead to the cropping of larger vehicles, which reduces accuracy. This paper proposes a method that combines the results of the YOLO model with an SAHI-enhanced YOLO version, after adjusting the confidence scores of bounding boxes using a Gaussian Mixture Model-based approach, to improve accuracy in detecting small vehicles without compromising the detection of larger objects. This combined approach enhances detection coverage by normalizing the model outputs based on synthesized data, including removing duplicates and score normalization, and then merging the two detection sets. We conduct a thorough evaluation of the proposed method with different versions of YOLO: YOLOv5, YOLOv6, YOLOv7 and YOLOv8. Experimental results on two aerial surveillance datasets, namely, Visdrone and DroneVehicle, demonstrate the increased accuracy of mAP across all tested models.

A Deep Spiking Neural Network (DSNN) with Heterogeneous Regularization learning technique is proposed to build a more biologically plausible approach that evaluates the amount of noise and finds a stopping criterion for fast realistic illumination. Our contribution is to introduce a model that improves the label propagation of DSNN and is more efficient on neuromorphic hardware than a corresponding Artificial Neural Network. More specifically, we develop a biological neural model with a heterogeneous regularization technique that works similarly to a human brain and can detect noise using deep spikes without relying on mathematical metrics to extract noise features. The objective function of the proposed DSNN consists of a supervised term and an unsupervised term. The supervised term enforces the matching term between the predicted labels and the known labels. The unsupervised term enforces the smoothness of the predicted labels of the entire data samples. By learning a DSNN with the proposed objective function, we are able to develop a more powerful learning algorithm. Experiments were conducted using scenes with Global Illumination and various image distortions. The proposed model was also compared with the human visual system and other state-of-the-art models. The results show better performance and advantages in terms of efficiency, an increasingly biologically plausible network, and ease of implementation in Neuromorphic Hardware.

With the breakthrough development of technology in the 4.0 digitalization era, computer vision and deep learning have emerged as promising technologies for industrial quality inspection. By leveraging the power of machine learning algorithms, computer vision systems can automatically detect and classify defects in industrial products with high precision and efficiency. As the system processes more data and identifies more complicated defects, it can become more accurate and efficient in detecting imperfections and ensuring product quality. This paper proposes an inspection system integrated with the YOLOv8 network to assess the quality of products based on their surface. The data multi-threading mechanism is also applied in the system to ensure real-time processing operations. The experimental results show that the proposed system reaches high detection accuracy among different types of defects, at above 90%. Additionally, the proposed model reveals that the scratch defect is the most difficult error to detect, requiring a long time for decision analysis.

The early detection of plant diseases by means of precise or automated detection techniques can improve the quality of food production and reduce economic losses. In this study, a deep convolutional neural network (Deep CNN) was devised to automatically and accurately identify leaf diseases in Cucumis Melo L. The deep CNN model used 1.776 sample images of healthy Cucumis Melo L. leaves and leaves infected with anthracnose, downy mildew, and powdery mildew for training and 198 random test images for the identification accuracy. We also conducted experiments comparing the identification results on our self-designed CNN model with three standard AlexNet, VGG16, and VGG19 models based on the same image dataset. The training experiment results with and without data augmentation while keeping the same parameters of the simulation showed that the accuracy of the self-designed CNN, AlexNet, VGG16, and VGG19 models were 93.58%, 90.26%, 92.86%, and 92.50%, respectively. The results confirmed the feasibility and effectiveness of our proposed CNN model for automatic leaf disease detection in the smart farming of Cucumis Melo L.

A traffic incident is one of the major concerns about traffic congestion on an urban expressway. Therefore, understanding the impact of traffic incidents can provide more accurate information to traffic management and participants. This study proposes a new approach that can be used to enhance the traffic incident impact in real situations. The approach combines the VISSIM simulation and machine learning algorithm. First, the fusion data were gathered from various sources, such as the global positioning system (GPS) data on the Seoul expressway, the traffic accident from the traffic accident analysis system (TAAS), and the traffic volume from the Seoul traffic information management system (TOPIS). Second, the speed, traffic volume, and accident information are matched together based on the national standard node link. Then, the VISSIM is used to analyze the speed change and expand incident data considering the incident severity. Finally, the long short-term memory (LSTM) is implemented for training the dataset and forecasting traffic incidents for the prediction model. The proposed approach could perform better accuracy prediction for the traffic incident impact to enhance the comfortable and safe operations of the urban expressway. Especially, the 10% VISSIM data penetration rate achieved the best performance with an average mean absolute percentage error (MAPE) of 24.80.

The curse of dimensionality is a problem that arises in datasets containing high-dimensional features. One of the primary methods used to address this issue is feature selection (FS), an effective technique for reducing insignificant features in a dataset. In this study, a two-stage approach called JASAL, which combines Filter and Wrapper FS methods, is proposed for high-dimensional medical datasets. In the first stage, preprocessing is applied to the high-dimensional datasets, and the ReliefF filter method is selected for this purpose. In the subsequent stage, continuous optimization algorithms, such as the Sine-Cosine Algorithm (SCA) and JAYA algorithms, are hybridized for binary optimization. A wrapper method that stochastically selects one of these two algorithms is used. Additionally, to overcome the local optimum problem of the JAYA algorithm, the Lévy flight strategy is incorporated into JAYA’s existing solution update strategy. To evaluate the performance of the JASAL algorithm, experiments are conducted on eight medical datasets, seven of which are high-dimensional. The results of the proposed algorithm are compared with 11 other metaheuristic algorithms. Various metrics, such as mean accuracy, mean number of selected features, and standard deviation, are considered in the experiments. The experimental results reveal that the JASAL algorithm achieves higher average accuracy than the compared algorithms. Furthermore, it is observed that these accuracy values are obtained by selecting relatively fewer features.

Many strategies have been devised to facilitate scheduling determinations per the dynamic production landscape. Also, the repercussions of environmental degradation intensify, and the focus on sustainable production methodologies has acquired substantial recognition. In addition, production managers are required to carefully determine the suitable schedule for each machine, considering environmental criteria. In such cases, another challenge in the manufacturing system alongside the minimization of job completion time which is important is reducing the environmental parameters impacts. Hence, this study focuses on addressing the challenge of distributed scheduling problems within the context of sustainable production. For energy-efficient factory selection, in this research, support vector machine (SVM) algorithm and improved particle swarm optimization (IPSO) have been used. In this case, first, factories with the least amount of pollution, waste, and energy consumption were selected. Second, the composition of intelligent algorithms such as gravitational search algorithm (GSA) and genetic algorithm (GA) was used to propose an appropriate distributed jobs scheduling in selected distributed factories. The simulation results show that this intelligent scheduling of distributed sustainable factories has significant potential to minimize environmental parameters inside production cycle optimization.

In the domain of deep learning, several architectures have been extensively used for time series data analysis and mining, including recurrent neural network (RNN)-based architectures such as gated recurrent units (GRUs) and long short-term memory (LSTM). Despite their effectiveness, these models face challenges in analyzing long-range dependencies due to their reliance on current time-step data. To address this limitation, the integration of convolutional neural networks (CNNs) into time-series analysis has been explored, enhancing model performance and accuracy. This paper focuses on addressing these challenges by applying a modified InceptionTime model for time-series classification (TSC). Our study conducts an in-depth review and evaluation of existing deep learning techniques for TSC, exploring potential model combinations to improve accuracy. Specifically, we investigate the integration of CNN and RNN models, leveraging their strengths for both TSC and image processing tasks. Our proposed model, evaluated in the UCR-85 data set, shows significant improvements in accuracy and performance, offering a promising approach for complex time series analysis problems.