AUTOMATIC CONTROL AND COMPUTER SCIENCES最新文献

A sun tracking system incorporated into a parabolic trough collector for precise control is presented in this study. The collector’s rotation axis is aligned with the east-west direction. With a concentration ratio of 160 and a narrow acceptance angle of 2 deg, achieving accurate tracking control is crucial for maximizing performance. To accomplish this, two established tracking configurations, namely open-loop and closed-loop, are utilized. The open-loop control utilizes a sun position algorithm. At the same time, the closed-loop system incorporates a sun sensor designed with light-dependent resistors. The proposed embedded system was verified using an experimental prototype. The experimental prototype was developed using the AVR ATMega32, a low-cost microcontroller. It was tested for tracking errors in both configurations. The outcome of the experimental prototype is presented in this work. The tracking controller provides a convenient solution to low-cost sun tracking using simple light-dependent resistors connected in a bridge configuration. The tracker’s accuracy can be conveniently controlled using the sun sensor’s threshold voltage, thus making it adaptable to different working environments.

Field crop pest and disease control knowledge texts contain rich core information such as pest and disease descriptions and control measures. However, it can be challenging to build a knowledge graph for field agricultural diseases due to certain domain characteristic, such as the use of specific terminology or pharmaceuticals, and multiple meanings of characters. Based on these analyses, we propose a named entity recognition method called Agri-NER-Net for field crop diseases and pests. The method firstly designs a multigranularity feature approach, combining characters, Chinese character glyphs, and words. Subsequently, we process these features using BiLSTM network pairs to model contextual long-range location-dependent features, and introduce a self-attention mechanism to enhance the model’s long-range dependency extraction capability. Finally, the LCRF (linear-conditional random field) model is used to predict the labelled sequence of target entities. The experimental results prove that the method proposed in this paper demonstrates a more excellent comprehensive recognition effect compared with the current mainstream named entity recognition models.

The desire to increase the energy of the pulses by which the impact excitation of ultrawideband antennas is carried out prompts some authors of equipment (radars) to change the shape of these pulses. In particular, pulses are made bipolar. This article explores the issue of optimizing the shift of the second component of the excitation pulse. Such a shift allows obtaining the maximum level of the spectrum at the frequency of interest, determined by the transmitter-receiver path.

Aiming to solve the low accuracy and slow speed of Chinese character recognition in the traditional license plate recognition, a method of license plate location, character segmentation and recognition using computer vision library OpenCV and license plate character recognition convolutional neural network (LPCR Net) is proposed. First, the RGB three-channel image is separated from the input image, and the input image is binarized by calculating the color characteristics of the license plate, then the multiple connected regions are obtained through morphological operations such as expansion and closure, the license plate location is completed via calculating the standard license plate aspect ratio and area; secondly, the horizontal and vertical projection method used in the traditional license plate character segmentation is improved to complete the license plate character segmentation, which improves the accuracy and speed of Chinese character segmentation; finally, the license plate character recognition is completed based on LPCR Net, and the recognition accuracy rate reaches 98.33%, which is 3.11% higher than that of AlexNet. Experimental results show that the proposed method can effectively improve the accuracy of license plate location, character segmentation and recognition.

A new prediction method based on improved attention mechanism and time convolutional network fusion is proposed for the prediction of groundwater level in karst areas. Within the overall framework of the prediction method, historical water level, flow rate, and rainfall were selected as input data. The input data is processed by the time attention module and the feature attention module respectively to form a weight matrix corresponding to the data sequence, and then trained and learned using a time convolutional network to complete prediction. Experimental results show that the proposed method is significantly better than LSTM method, RNN method and CNN method in terms of mean absolute error and root-mean-square deviation. The predicted change curves at the three measurement points also form a good agreement with the actual groundwater level change curve.

Intelligent transport systems (ITSs) have emerged as a groundbreaking solution to address the challenges associated with road traffic, which are enhancing road utilization efficiency, providing convenient and safe transportation, and reducing energy consumption. ITS leverages advanced technologies to collect, store, and deliver real-time road traffic information, enabling intelligent decision-making and optimizing various aspects of transportation systems. As a contribution in this matter, we propose in this paper a novel efficient macroscopic approach, based on the multihead self-attention vision transformer (MSViT), for categorizing road traffic congestion, from nighttime videos, into three classes: light, medium, and heavy. To assess the performance of our approach, we conducted experiments using the nighttime UCSD (University of California San Diego) dataset, which includes various weather conditions (clear, overcast, and rainy) and traffic scenarios (light, medium, and heavy). The classification accuracy reached a high level of 94.24%. By incorporating a support vector machine (SVM) classifier into this method, we managed to enhance this accuracy to the outstanding level of 98.92%, thus outperforming the existing state-of-the-art methods that were evaluated using the same UCSD dataset, furthermore, the execution time was optimized.

With the construction of smart grid, aerial insulator defect detection based on computer vision has become an important task to ensure grid safety. When the target detection model is too large, it is not conducive to the edge deployment of aerial inspection UAV; Moreover, different aerial photography angles and distances will cause the insulator string in the image to have any direction and less defect information. In order to solve these problems, this paper proposes a rotating GBS-AFP-YOLOv5 model with the combination of lightweight and adaptive features. Firstly, an improved YOLOv5 based on lightweight GBS is proposed by Ghost convolution, which can effectively extract features while reducing the complexity of the model. Then, an adaptive information interaction feature pyramid (AFP) is proposed by combining CARAFE upsampling operator and ECA attention, which effectively fuses the feature information of shallow and deep defects and improves the model performance. Then, a more accurate insulator string detection method is realized by using rotating frame combined with ring label smoothing technology. Finally, the normalized wasserstein distance (NWD) is introduced to improve the loss function, which further enhances the detection ability of the model for small targets with defects. Based on the insulator data set, the test results show that the model has a good defect detection performance, which is improved from mAP0.5 to 0.923 on the basis of only 4.32M parameters.

The evolution of recommendation systems has revolutionized user experiences by providing personalized recommendations. Although conventional systems such as collaborative and content-based filtering are reliable, they still suffer from inherent limitations. We introduce a hybrid recommendation system that combines content-based filtering using TF-IDF and cosine similarity with collaborative filtering and SVD to address these challenges. We bolster our model through supervised machine learning algorithms like decision trees (DT), random forests (RF), and support vector regression (SVR). To amplify predictive prowess, boosting algorithms including CatBoost and XGBoost are harnessed. Our experiments are performed on the benchmark dataset MovieLens 1M, which highlights the superiority of our hybrid method over more traditional alternatives with SVR being the best-performing algorithm consistently. Our hybrid model achieved an MSLE score of 2.3 and an RMSLE score of 1.5, making SVR consistently the best-performing algorithm in the recommendation system. This combination demonstrates the potential of collaborative-content hybrids supported by cutting-edge machine-learning techniques to reshape the field of recommendation systems.

A fusion optimization algorithm has been proposed to enhance the reliability and accuracy of regular surface feature extraction from laser point clouds. to get optimal result. Firstly, the Octree-based constrained adaptive growth method is utilized to optimize the neighborhood points of point cloud and establish its topological relationship. Secondly, the Harris-3D algorithm is applied to extract key points from the point cloud data, followed by a region growth method that combines double thresholds of normal vector angle and Euclidean distance, to segment the point cloud into separate clusters. Finally, regular surface features are extracted from these clusters, allowing for the recognition of 3D object surface morphology and features. Experiments on regular surface feature extraction from point clouds have shown that the proposed fusion optimization algorithm can significantly improve the accuracy and efficiency of feature extraction. The RMS errors for the extraction and reconstruction of quadric surfaces like planes, cylinders, cones, and spheres are below 0.020 mm. Additionally, a real-world experiment involving a large amount of complex point cloud data from an unmanned laser scanning scene also confirms the effectiveness of the proposed feature extraction optimization algorithm for regular surface feature extraction, object recognition, and 3D reconstruction.

The agriculture sector is the backbone of every country. The growth of a country is complete only if there is an increase in agricultural products following the increase in population. But this ratio is often not maintained due to climate change and pest attacks causing huge crop damage. Therefore, a large amount of pesticides and chemicals are used in agriculture today. Massive chemicals application not only affects the crops but also the air. The use of chemicals has a large impact on air pollution, which causes respiratory diseases and various types of allergies. Therefore, a method is needed to detect these chemicals in the air in real-time. Here proposes an IoT-based system that uses two sensors to measure concentration levels of different harmful chemicals and two machine learning algorithms logistic regression, and support vector machine (SVM) to predict the risk of air pollution. Using the sensed data, the system calculates the air quality index (AQI). The proposed system will be very useful for officials as well as common people to find the quality of air in a particular area.