Intrusion detection, as a technology used to monitor abnormal behavior and maintain network security, has attracted many researchers’ attention in recent years. Thereinto, association rule mining is one of the mainstream methods to construct intrusion detection systems (IDS). However, the existing association rule algorithms face the challenges of high false positive rate and low detection rate. Meanwhile, too many rules might lead to the uncertainty increase that affects the performance of IDS. In order to tackle the above problems, a modified genetic network programming (GNP) is proposed for class association rule mining. Specifically, based on the property that node connections in the directed graph structure of GNP can be used to construct attribute associations, we propose to introduce information gain into GNP node selection. The most important attributes are thus selected, and the irrelevant attributes are removed before the rule is extracted. Moreover, not only the uncertainty among the class association rules is alleviated and also time consumption is reduced. The extracted rules can be applied to any classifier without affecting the detection performance. Experiment results based on NSL-KDD and KDDCup99 verify the performance of our proposed algorithm.

In this study, we present an optimal control model for an uncertain production inventory problem with deteriorating items. The dynamics of the model includes perturbation by an uncertain canonical process. An expected value optimal control model is established based on the uncertainty theory. The aim of this study is to apply the optimal control theory to solve a production inventory problem with deteriorating items and derive an optimal inventory level and production rate that would maximize the expected revenue. The uncertainty theory is used to obtain the equation of optimality. The Hamilton–Jacobi–Bellman (HJB) principle is used to solve the equation of optimality. The results are discussed using numerical experiments for different demand functions.

To complete the wiring operation of the main transmission line stripped of its insulating skin in a live power distribution system, a structured-light binocular vision method is utilized to identify and locate the line. First, aiming at the interference of the background information, a depth threshold segmentation method is used to filter the background area. Second, a mean filtering method is proposed to filter out the mismatch noise of a binocular vision camera in an outdoor environment. The Canny algorithm is then utilized to extract the contour, the central axis of the main transmission line is fitted, and the difference in the neighborhood pixel value is used to recognize the stripping area. Finally, the spatial equation and attitude of the central axis of the fitting transmission line are obtained along with the central coordinates of the stripping area, guiding the robot to carry out the wiring.

Blood type identification is a method used for determining the specific blood type of a person. It is a requirement before blood transfusions or blood donations is undertaken especially during emergency situations. Presently, the tests are performed manually by medical technologists in the laboratories. Sometimes, manual blood typing is prone to human error, resulting to incorrect blood grouping and wrong typing in the report, leading to fatal transfusion reactions. The study was focused on the development of a device that is capable of identifying the blood type of an individual using an image processing and machine learning algorithms. The study covered the identification of eight blood types, specifically rhesus positive and negative, A, B, O, and AB, by developing a capturing box integrated with a web camera system that could effectively capture blood sample images. In this study, the methodologies utilized were image processing through segmentation, feature extraction by color and texture properties, and different machine learning algorithms. After training, the results showed that coarse tree DT has the best performance accuracy score of 97.77% using 70:30 holdout validation. The testing results showed that the system is 100% accurate as validated by a registered medical technologist.

Robots make decisions in a variety of situations requiring multitasking. Therefore, in this work, a method is studied to address multiple tasks based on reinforcement learning. Our previous method selects an action when the q-values of the action for each task correspond to a priority value in the q-table. However, the decision-making would select an ineffective action in particular situations. In this study, an action value weighted by priority is defined (termed as action priority) to indicate that the selected action is effective in accomplishing the task. Subsequently a method is proposed for selecting actions using action priorities. It is demonstrated that the proposed method can accomplish tasks faster with fewer errors.

Internet of Things (IoT) is a highly integrated and comprehensive information technology that is currently a research hotspot. However, it presents many challenges; e.g., the use of multiple products in one IoT can lead to incompatible communication. It is imperative to solve the problem of communication protocol heterogeneity. In this study, a multi-protocol fusion IoT called MPFIoT was designed, implemented, and applied to environmental monitoring in a community. A hierarchical design was adopted in MPFIoT. It was divided into four layers: a data perception layer, a wireless transmission layer, a gateway layer, and an application layer. First, in the data perception layer, various sensors were used in the community to collect diverse environmental information. Second, in the wireless transmission layer, four wireless communication protocols were applied: Wi-Fi, LoRa, ZigBee, and Bluetooth. Third, the gateway layer had two sub-layers: the protocol fusion sub-layer was designed to reduce the degree of heterogeneity between protocols, and the control command sub-layer was used to control nodes via the control command frame. Finally, the application layer communicated with the gateway layer through the TCP/IP protocol. Thus, MPFIoT with four layers was developed, and its functions and performance were tested. The test results indicated that multiple types of environmental data can be collected and transmitted by MPFIoT. The packet loss ratio was less than 2% and the transmission delay was approximately 500 ms, which satisfied the application requirements.

Surface electromyography (sEMG) signals play an essential role in disease diagnosis and rehabilitation. This study applied a powerful machine learning algorithm called extreme gradient boosting (XGBoost) to classify sEMG signals acquired from muscles around the knee for distinguishing patients with knee osteoarthritis (KOA) from healthy subjects. First, to improve data quality, we preprocessed the data via interpolation and normalization. Next, to ensure the description integrity of model input, we extracted nine time-domain features based on the statistical characteristics of sEMG signals over time. Finally, we classified the samples using XGBoost and cross-validation (CV) and compared the results to those produced by the support vector machine (SVM) and the deep neural network (DNN). Experimental results illustrate that the presented method effectively improves classification performance. Moreover, compared with the SVM and the DNN, XGBoost has higher accuracy and better classification performance, which indicates its advantages in the classification of patients with KOA based on sEMG signals.

The visual SLAM system requires precise localization. To obtain consistent feature matching results, visual features acquired by neural networks are being increasingly used to replace traditional manual features in situations with weak texture, motion blur, or repeated patterns. However, to improve the level of accuracy, most deep learning enhanced SLAM systems, which have a decreased efficiency. In this paper, we propose Coarse TRVO, a visual odometry system that uses deep learning for feature matching. The deep learning network uses a CNN and transformer structures to provide dense high-quality end-to-end matches for a pair of images, even under indistinctive settings with low-texture regions or repeating patterns occupying the majority of the field of view. Meanwhile, we made the proposed model compatible with NVIDIA TensorRT runtime to boost the performance of the algorithm. After obtaining the matching point pairs, the camera pose is solved in an optimized way by minimizing the re-projection error of the feature points. Experiments based on multiple data sets and real environments show that Coarse TRVO achieves a higher robustness and relative positioning accuracy in comparison with the current mainstream visual SLAM system.

The terahertz (THz) region has much appeal for differentiating between hydrate systems and for physically characterizing pharmaceutical drug materials. The present study employs THz absorption spectroscopy to investigate the effect of heating on dehydration and hydration in α-lactose monohydrate. Distinctive THz absorption spectra were observed following various heating durations. The THz absorption spectra for α-lactose monohydrate and anhydrous α-lactose exhibit clear differences. Pure α-lactose monohydrate displays clear absorption peaks at 0.53, 1.05, 1.11, 1.33, and 1.56 THz. The complete dehydration of α-lactose monohydrate takes only 15 mins at 145°C (418 K). Moreover, the THz refractive index of α-lactose monohydrate decreases during dehydration. The dehydration of α-lactose monohydrate was also studied using Beer–Lambert law to compare THz absorption spectra as functions of the heating time. The absorption coefficient spectra recorded at 0.53 and 1.35 THz for α-lactose monohydrate after different dehydration times vary linearly with the remaining water content.

Falling is a major challenge faced by elderly and disabled people who live alone. They therefore need reliable surveillance so they can be assisted in the event of a fall. An effective fall detection system is needed to provide good care to such people as it will allow for communication with caregivers. Such a system will not only reduce the medical costs related to falls but also lower the death rate among elderly and disabled people due to falls. This review paper presents a survey of different fall detection techniques and algorithms used for fall detection. Various fall detection approaches including wearable, vision, ambience, and multimodal systems are analyzed and compared and recommendations are presented.

Edge cloud is used to handle latency-sensitive services. However, due to the large number of concurrent requests for edge intensive tasks, the resource allocation strategy affects the stability of nodes. In addition to an adaptive resource allocation model based on chaotic hierarchical gene replication (CRPSO model), the concept of chaotic replication ratio is proposed. This study is divided into two parts. The first is to verify the algorithm verification of the simulation platform. By comparison, it is found that CRPSO reduces the CPU and bandwidth utilization by 43.7% and 62.7% on average, respectively, and the memory usage is also lower than other algorithms. Thereafter, we compared the CRPSO algorithm with the Kubernetes clustering algorithm. Experiments showed that the fitness of the CRPSO model is 33.7% higher than that of the comparison algorithm on average. The algorithm is superior to the cluster scheduling algorithm in terms of CPU utilization and memory utilization. Furthermore, the total variance of the two resources involved in this model improved significantly, reaching 69.8% on average. In addition, CRPSO also has great advantages in other aspects of CPU and memory. It is indicated that the model in this study is suitable for the scenario of edge large-scale requests.

An event-triggered network predictive control method, which uses allowable time delays, was developed for networked control systems with random network delays, packet disorders, and packet dropouts in the feedback and forward channels. In this method, random communication constraints are uniformly treated as a time delay at each time instant. Subsequently, based on a time-delay state feedback control law, the proposed method is used to actively compensate for the time delay that exceeds the allowable. In addition, the introduction of an event-triggered mechanism reduces communication loads and saves network resources. A necessary and sufficient stability condition for the closed-loop system is provided, which is independent of random time delays and is related to the allowable delay. Finally, the simulation results of the two systems verified the effectiveness of the proposed method.

This paper proposes a transfer learning method for an object detection model using a genetic algorithm to solve the difficulty of the conventional transfer learning of deep learning-based object detection models. The genetic algorithm of the proposed method can select the re-learning layers automatically in the transfer learning process instead of a trial-and-error selection of the conventional method. Transfer learning was performed using the EfficientDet-d0 model pre-trained on the COCO dataset and the Global Wheat Head Detection (GWHD) dataset, and experiments were conducted to compare fine-tuning and the proposed method. Using the training data and the validation data of the GWHD, we compare the mean average precision (mAP) of the models trained by the conventional and the proposed methods, respectively, on the test data of the GWHD. It is confirmed that the model trained by the proposed method has higher performance than the model trained by the conventional method. The average of mAP of the proposed method, which automatically selects the re-learning layer (≈0.603), is higher than the average of mAP of the conventional method (≈0.594). Furthermore, the standard deviation of results obtained by the proposed method is smaller than that of the conventional method, and it shows the stable learning process of the proposed method.

This paper estimates the state transition of COVID-19 positive cases by analyzing the data about confirmed positive cases in Tokyo, Japan. The prediction of the number of newly infected persons is one of the active research topics for the COVID-19 pandemic. Although such a prediction is important for recognizing the future risk of spreading infectious diseases, understanding the state transition after they are confirmed to be positive is also important for estimating the number of required ICUs, hotel rooms for isolation, etc. This paper classifies the state after being positive into “in hotel/home for isolation,” “in hospital with a mild state,” “in hospital with a severe state,” “recovered,” and “dead” and estimates the transition probabilities among those states from the data about confirmed positive cases in Tokyo, Japan. This paper shows the parameters estimated from different periods and discusses the difference considering the pandemic situation. An agent simulation using the estimated transition probabilities as its parameters is also proposed. The result of the simulation from August to November 2020 shows the predicted number of agents is close to the actual data. As one of the possible applications to the proposed agent simulation, this paper shows the simulation result from December 2020 to January 2021 under a hypothetical situation.

China is currently in a new phase of transition from high-speed growth to high-quality growth, and accurate estimation of element outputs is essential for the smooth progress of the transition. Using the back-fitting method, this study constructed a model of a spatiotemporal-varying elasticity production function to estimate the factor-output elasticity from 1993 to 2017 in 31 Chinese provinces. Nonparametric kernel density method was applied to describe the spatiotemporal evolution characteristics of factor-output elasticity. The results show that the factor-output elasticity of different provinces shows a nonlinear change trend over time and between regions. Overall, the elasticity of labor output shows a decreasing trend, the elasticity of capital output shows an increasing tendency, the eastern region has the lowest level of labor-output elasticity, but the highest level of capital-output elasticity. The western region has the highest level of labor-output elasticity but the lowest level of capital-output elasticity. On the whole, regions with higher resilience in labor output gradually shift towards the West, while capital shifts towards the East.

The scientific analysis of the slalom training process can significantly improve the performance of athletes. In this paper, the P matrix is defined by extracting the multi-joint space coordinate trajectories of the athletes in the video to analyze the slalom training pattern. The principal component analysis was used to extract the main eigenvalues and eigenvectors of the P matrix, which were defined as the main eigenbehaviors of slalom skiing, and six main eigenbehaviors were used to achieve a similarity of 96% between the reconstructed skiing sequence and the original sequence. Similarly, the group characteristic S matrix is constructed by using the individual eigenbehaviors, and the eigenvectors of the matrix are used to define the characteristic behavior of the group to classify the hierarchical group and determine the group to which the individual belongs. Results show that this method can better identify the movement pattern of the human body’s multi-joint space trajectory in indoor or outdoor slalom skiing, and provide scientific guidance for skiing training, so that athletes can achieve better training effectiveness.

To achieve healthy development and optimal growth for harvest in an aquaculture system, correct determination of fish growth stages is very important. The sizes or growth stages of the fish are used by farm managers to regulate stocking densities, optimize daily feeding, and ultimately choose the ideal time for harvesting. This paper presented a vision system-based fish classification using pixel transformation and neural network pattern recognition. Morphometrics parameters are used to facilitate a supervised gathering of datasets. Before feature extraction, the images go through intensity transformation using histogram analysis and Otsu’s thresholding. Using Pearson’s correlation coefficient, the six most important characteristics of the original ten attributes were identified. The developed intelligent model using neural network pattern recognition has an overall training accuracy equal to 90.3%. The validation, test, and overall accuracy are equal to 85.7%, 85.7%, and 88.9%, respectively.

Real-time water quality index (WQI) monitoring – a simplified single variable indication of water quality (WQ) – is vital in attaining a sustainable future in precision aquaculture. Although several monitoring systems for water quality parameters (WQP) use IoT, there is no existing WQI IoT monitoring for Oreochromis niloticus because the current WQI models are too complex to be deployed for low-level computing platforms such as the IoT modules and dashboards. Thus, the development of the IoT-based WQI fuzzy inference system (FIS) was simplified by the multi-gene genetic programming (MGGP) to search for non-linear equations given the simulated WQP fuzzy sets. Results have shown that the implemented novel system can accurately predict the WQI IoT monitoring with an average of R² and RMSE of 0.9112 and 0.6441, respectively. Implementing WQI in the IoT monitoring dashboard using the MGGP has significantly addressed the present challenges in deploying other complex AI-based models for WQI, such as the FIS and neural networks in low-computing capable platforms.

Lithium iron phosphate (LiFePO₄) has become the top choice battery chemical in photovoltaic (PV) system nowadays due to numerous advantages as compared to lead acid batteries. However, LiFePO₄ needs a battery management system to optimize energy utilization. State of charge (SoC), state of health (SoH), cell balancing, remaining useful life are some of its crucial parameters. This review paper discusses overview of battery management system (BMS) functions, LiFePO₄ characteristics, key issues, estimation techniques, main features, and drawbacks of using this battery type.

The safety and security of an individual is important in our society. Bombing attacks can cause significant destruction and death. Energy efficient and compact bomb removal robots are challenging to develop because these typically involved a large array of sensors individually acquiring gas data. This study addresses this challenge by developing a multiple bomb-related gas prediction model using machine learning and the electronic nose sensor substitution technique. Three models can predict gasses such as ammonia, ethanol, and isobutylene using only carbon monoxide, toluene, and methane sensors. The feedforward artificial neural network (FFNN) with three hidden layers was optimized for the regression of each target gas. Consequently, ammonia, ethanol, and isobutylene predictions achieved R² values of 1, 1, and 1 as well as MSE values of 0.35696, 0.052995, and 0.0022953, respectively. This study demonstrates that the sensor substitution model (BombNose) is highly reliable and appropriately sensitive in the field of bomb detection.

The primary purpose of this research is to implement Deeplabv3 architecture’s deep neural network in detecting and segmenting portable X-ray source model parts such as body, handle, and aperture in the same color scheme scenario. Similarly, the aperture is smaller with lower resolution making deep convolutional neural networks more difficult to segment. As the input feature map diminishes as the net progresses, information about the aperture or the object on a smaller scale may be lost. It recommends using Deeplabv3 architecture to overcome this issue, as it is successful for semantic segmentation. Based on the experiment conducted, the average precision of the body, handle, and aperture of the portable X-ray source model are 91.75%, 20.41%, and 6.25%, respectively. Moreover, it indicates that detecting the “body” part has the highest average precision. In contrast, the detection of the “aperture” part has the lowest average precision. Likewise, the study found that using Deeplabv3 deep neural network architecture, detection, and segmentation of the portable X-ray source model was successful but needed improvement to increase the overall mean AP of 39.47%.

The understanding of vision-based data acquisition and processing aids in developing predictive frameworks and decision support systems for efficient aquaculture monitoring and management. However, this emerging field is confronted by a lack of high-quality underwater visual data, whether from public or local setups and high cost of development. In this regard, an underwater camera that captures underwater images from an inland freshwater aquaculture setup was proposed. The components of the underwater camera system are primarily based on Raspberry Pi, an open-source computing platform. The underwater camera continuously provides a real-time video streaming link of underwater scenes, and the local processor periodically acquires and stores data from this link in the form of images. These data are stored locally and remotely. Based on the results of the developed low-cost underwater camera, it captures and differentiate fish region to its background before and after flushing as influenced by turbidity. Hence, the developed camera can be used for both aquarium and inland aquaculture pond setup for fish monitoring.