Search

This study focuses on the problem of multitarget tracking. To address the existing problems of current tracking algorithms, as manifested by the time consumption of subgroup separation and the uneven group size of unmanned aerial vehicles (UAVs) for target tracking, a multitarget tracking control algorithm under local information selection interaction is proposed. First, on the basis of location, number, and perceived target information of neighboring UAVs, a temporary leader selection strategy is designed to realize the local follow-up movement of UAVs when the UAVs cannot fully perceive the target. Second, in combination with the basic rules of cluster movement and target information perception factors, distributed control equations are designed to achieve a rapid gathering of UAVs and consistent tracking of multiple targets. Lastly, the simulation experiments are conducted in two- and three-dimensional spaces. Under a certain number of UAVs, clustering speed of the proposed algorithm is less than 3s, and the equal probability of the UAV subgroup size after group separation is over 78%.

In the field of intelligent robot engineering, whether it is humanoid, bionic or vehicle robots, the driving forms of standing, moving and walking, and the consciousness discrimination of the environment in which they are located have always been the focus and difficulty of research. Based on such problems, Naive Bayes Classifier (NBC), Support Vector Machine(SVM), k-Nearest-Neighbor (KNN), Decision Tree (DT), Random Forest (RF) and eXtreme Gradient Boosting (XGBoost) were introduced to conduct experiments. The six individual classifiers have an obvious effect on a particular type of ground, but the overall performance is poor. Therefore, the paper proposes a “Novel Hybrid Evolutionary Learning” method (NHEL) which combines every single classifier by means of weighted voting and adopts an improved genetic algorithm (GA) to obtain the optimal weight. According to the fitness function and evolution times, this paper designs the adaptively changing crossover and mutation rate and applies the conjugate gradient (CG) to enhance GA. By making full use of the global search capabilities of GA and the fast local search ability of CG, the convergence speed is accelerated and the search precision is upgraded. The experimental results show that the performance of the proposed model is significantly better than individual machine learning and ensemble classifiers.

Under the condition of a strong electric field, partial discharge often occurs when insulated wire is damaged. The recognition of partial discharge is an effective method for the fast and accurate detection of high voltage insulated wire faults. This paper proposes a PD recognition algorithm based on a convolutional neural network and long short-term memory (LSTM). In addition, attention mechanisms are introduced to give separate weights to LSTM hidden states through a mapping, weighting, and learning parameter matrix. This is done to reduce the loss of historical information and to strengthen the influence of important information. The complex relationship between the voltage signal change and the grid operation state response has been established. The proposed method is verified by the ENET data set published by VSB University. The recognition accuracy is 95.16% for no-PD and 94.44% for PD. Results from the proposed algorithm show that this method has a higher detection accuracy.

Insulated conductors can improve the stability of power transmission and reduce the construction space compared with traditional bare conductors. Therefore, insulated conductors are used more and more in overhead power transmission. However, a major challenge of using insulated overhead conductors (IOC) is that the ordinary protection devices are not able to detect the phase-to-ground faults and something, such as tree branch, hitting conductor events. This may cause an accident such as power failure or electrical fire and result in serious damage. In this paper, a new approach, which is based on Discrete Wavelet Transform (DWT) and Long Short Term Memory network (LSTM) for detecting of IOC fault according to partial discharge, is presented. Firstly, the original signal is denoised by DWT. Secondly, the denoised signal is decomposed and extracted features on different layers by DWT. Finally, IOC fault is detected by LSTM. This method can improve the detection accuracy of IOC fault which is tested on the ENET public data set and compared with other classification methods.

In order to quickly and effectively detect lung information in different medical images, this paper designs an improved VGG16 image-based lung opacity classification detection method based on deep transfer learning. This paper applies offline data enhancement technology to increase the number of samples, improves VGG, and employs transfer learning to train the lung recognition model. The results show that the improved VGG16 network has an accuracy rate of 85% for the classification and recognition of lung pictures, and can accurately detect lung pathological mutation information.

Road traffic prediction plays a remarkable role in intelligent transportation system managements. Most current research usually considers the two dimensions of time and space dependencies. However, merely considering the upstream and downstream information in the spatial dimension is incapable to capture sufficient spatial features. In this paper, a traffic prediction model using Spatial-Temporal information and Traffic Pattern Similarity information (STTPS) is proposed. It selects observation points with similar traffic pattern to supplement spatial information through similarity analysis and attention mechanism. Simultaneously, many influences of seasonality and super-recent factors are fully considered in the time dimension. Furthermore, a time convolution network and a gated recurrent unit are used for feature extraction. Finally, the time characteristics of variant scales is fused to obtain the prediction result. The proposed prediction is compared with the state-of-art models to demonstrate its capability in both single-step (5 min) and multi-step (1 hour) traffic forecasting.

Aiming at some usage scenarios that have strict requirements on the scale of fusion operation, this paper proposes a lightweight infrared and visible image fusion algorithm. The algorithm first uses bilateral filtering to decompose the image into the detail layer and the base layer. Secondly, the fusion of detail layers is achieved through regional variance. Then, the LC saliency detection method is performed on the original image to generate a saliency map. A weight map is generated from the saliency map of the original image to guide the base layer to perform fusion. Finally, the fusion image is obtained through linear transformation of the detail layer and the base layer. The proposed algorithm architecture is simplified, and only bilateral filtering is used to realize multi-scale decomposition, which reduces the operation scale. In terms of the fusion quality, after experimental verification, the proposed algorithm has certain advantages in the clarity and edge transition of the fused image compared with the comparison algorithm.

A single swarm intelligence algorithm will be limited by its own search ability, and when dealing with some more complex matters, there will be problems such as low accuracy of solution and easy to fall into local optimum. However, the hybrid optimization method can achieve the purpose of improving the performance of the algorithm through information exchange between different populations. Therefore, in this paper, the particle swarm algorithm has the characteristics of individual memory, and the particle position update is used to replace the gray wolf individual position update, so that the gray wolf algorithm has memorability in the optimization. It can be seen from the optimization results of eight benchmark test functions that the "Multiply Species Cooperative Evolution" method combines the advantages of different algorithms, and it has obvious global convergence and strong competitiveness.

LightGBM is a fast distributed, high-performance gradient boosting framework based on decision tree algorithm, which can be used for sorting, classification, regression and many other machine learning tasks. In order to realize the classification of the terrain environment where the robot is located, this paper uses the gradient boosting tree LightGBM to identify and classify the IMU signals of the robot feet. First of all, the quaternion Euler transform and segmentation are performed on the signal. Secondly, the different period segments of the signal are analyzed and feature fused from the time domain and frequency domain respectively. Finally, the decision tree integrated hybrid classifier with the best parameters, namely LightGBM, is employed to classify the robot on the ground. The experimental results show that the accuracy of this method for the recognition of robot IMU signals on different surfaces reaches 90.7%.

With the continuous development of sensor technology, computer technology, artificial intelligence and other advanced technologies, there are more and more researches on trajectory tracking and detection technology, which have been widely used in urban planning, traffic management, safety control and other aspects. Trajectory tracking and detection has always been the focus of research by experts and scholars. The purpose of this study is to track and detect the spatial trajectory of the infected person under the current new crown virus epidemic, to timely and accurately understand the itinerary of the new crown virus infected person and to find out all the suspected contacts that the infected person may come into contact with. The current epidemic situation in various countries has made a certain contribution.

In the global fight against the novel corona-virus pneumonia epidemic (COVID-19), a reasonable prediction of the spread of the epidemic has important reference significance for epidemic prevention and control. In order to solve the problem of time series prediction and analysis of the epidemic with limited sample data, nonlinear and high-dimensional features, this study applies the Nonlinear Auto-Regressive neural network (NAR) model for machine learning. The paper predicts the development of the epidemic in the two dimensions of the number of confirmed cases and the number of deaths in major countries in the world, and compares NAR with the traditional Logistic Regression (LR), the classic time series model ARIMA and the SEIR infectious disease dynamic model. This research provides rapid decision-making and new ideas for countries to respond to the “post-epidemic era”.

Intelligent prediction and pattern recognition (PR) classification technology based on deep learning has attracted widespread attention from scholars in the field because it does not rely on subjective human analysis and reasoning. In this paper, a variant of convolutional neural network and recurrent neural network (RNN) is constructed by combining feature engineering and deep learning. A convolution-gated recurrent unit network (i.e., CONV-GRU) is proposed by improving the network structure and optimization algorithm. The model has an accurate effect on the recognition of the contact surface of autonomous mobile robots, with an accuracy rate of 90%. In addition, CONV-GRU also solves the over-fitting phenomenon of the benchmark model (RNN, LSTM, GRU). The research aims to solve the key problems of intelligent robot in tactile sensing, signal processing and intelligent computing technology.

When a series arc fault occurs in indoor power distribution system, current value of circuit is often less than the threshold of the circuit breaker, but the temperature of arc combustion can be as high as thousands of degrees, which can lead to electrical fire. The arc fault experimental platform is used to collect circuit current data of normal work and arc fault. Five types of loads which are commonly used in indoor distribution system, such as resistive and inductive in series load, resistive load, series motor load, switching power load and eddy current load, are chosen. This paper uses four features of current in time domain, i.e. current average, current pole difference, difference current average and current variance. Ten features of current in frequency domain are extracted by Fast Fourier Transform (FFT). The learning vector quantization neural network (LVQ-NN) is designed to judge the load type. The support vector machine optimized by particle swarm optimization (PSO-SVM) is designed to detect the arc fault. The simulation results show the effectiveness of the proposed method.

In order to improve the accuracy of mobile robot ground classification, a robot ground recognition method based on convolutional neural network (CNN) and long short-term memory neural network (LSTM) is proposed. The method takes historical collected data as input, builds a CNN architecture composed of one-dimensional convolutional layer and pooling layer, and extracts high-dimensional features reflecting the complex dynamic changes of the robot’s walking state. The CNN architecture constructs the proposed feature vector into a time series form as the input of the LSTM network and models the dynamic changes of the learning features. Taking the public data set in the CareerCon 2019 competition as experimental data, the results show that the algorithm in this paper has achieved better classification results on 9 types of ground. Comparative experiments with other advanced methods show that this method can further improve the classification accuracy of the robot on the ground, and provide technical support for realizing more effective ground environment perception and navigation control operation in the best state.

Under the background of the booming development of big data and the robot industry, the existing navigation technology has a certain delay in the identification of robot drive form and the unstructured surface physical environment. In this paper, by processing the statistical data of the robot foot sensor, the hybrid features of the robot in different environments are constructed in the two dimensions of time domain and frequency domain (fast Fourier transform and continuous wavelet transform). Finally, by combining decision tree and random forest algorithm, the mixed features are fully learned, and the high-precision classification results are obtained.

In recent years, with the third wave of development of artificial intelligence (AI), multi-agent system has become a hot topic in the field of distributed AI. As an important form of intelligent operation, multi-intelligence UAV cluster operation is emerging. This paper establishes a “Red vs Blue Multi-UAVs Cooperative Confrontation Strategy” (RB-MCCS). According to the red and blue UAV penetration and interception problems, this paper designs the UAV breakout cost and the optimal strategy optimization models. Furthermore, a swarm intelligence optimization method is introduced to solve the model.

In the context of the new era, the concept of smart transportation has appeared in people’s lives. Detecting vehicles and pedestrians has become a popular application research direction in the field of target detection. Aiming at the problem that traditional methods have low accuracy in vehicle detection in the actual environment, a vehicle detection method based on the CenterNet model in deep learning is proposed. When constructing the model, this paper regards the target as a point— the center point of the target BBox. The detector uses key point estimation to find the center point and returns to other target attributes, such as size, 3D position, direction, and even pose. This paper uses Peking University/Baidu_Autonomous Driving dataset for training and testing. The experimental results show that compared with Inception-ResNet-V2 and Efficient-Det, the method proposed in this paper has significantly improved the average detection accuracy. It has a good detection effect for vehicles in actual scenes, and the network has certain robustness.

Under the wave of the big data era, we need to analyze some data sets that cannot be captured and processed with conventional software tools within a certain time frame with new processing modes, so as to obtain stronger decision-making power and insight. In view of this, this paper uses a big data pattern matching algorithm to establish a SOM network (Self-Organizing feature Map) to calculate the probability of auto insurance renewal under the background of complex data. On the one hand, this research will increase the insurance renewal rate. In addition, it can provide advice and reference for the cultivation of auto insurance customer loyalty under the epidemic situation, and has a certain reference role for the development of my country’s insurance digitalization.

Deep learning is becoming more and more popular to solve image segmentation tasks in medical image processing community because of the incredible advantages in deep feature representation and nonlinear problem modeling. However, most existing deep learning methods based segmentation are implemented by combing deep, semantic, coarse-grained feature maps from the decoder sub network with shallow, low-level, fine-grained feature maps from the encoder sub-network, which are not up to the mustard of medical image segmentation. To solve the above-mentioned problem, an innovative end-to-end Pneumothorax Segmentation (PS) method based on UNet++ is proposed, where change maps could be learned from scratch using existing annotated datasets. And the fusion strategy of multiple side outputs is applied to combine change maps from different semantic levels. The high efficiency and availability of our proposed method are proved with SIIM-ACR Pneumothorax Segmentation dataset. Plenty of experimental results have shown that our proposed approach outperforms many cutting-edge methods.

Aiming at the problem that traditional flower classification methods and ordinary convolutional neural networks are difficult to reduce the effect of flower background, the classification effect is not ideal. This paper designs a flower classification model that combines saliency detection and VGG-16 convolutional neural network, and adopts stochastic gradient descent algorithm and prevents over-fitting technology to improve the model. Experiments on the international public flower recognition data set Oxford flower-102 show that the model proposed in this paper is better than other traditional network models and has high recognition accuracy, robustness and generalization ability, which can classify flowers and have higher practical value accurately and quickly.

Swarm evolution learning is to design machine learning algorithms based on optimization tools provided by evolutionary algorithms. Evolutionary learning is also called evolutionary algorithms. It is an algorithm cluster inspired by the biological evolution of nature. With the impact of global warming and climate change in recent years, a large number of cold-water fishes have been migrating, and the planning and prediction of migration paths has become a major research hot-spot. In this paper, a semi-supervised clustering neural network is used to build a model to predict the most likely migration direction of fish in Scotland, and a simulated annealing algorithm is used to calculate the maximum profit. This study will have certain implications for the development of fisheries and the allocation and planning of future marine resources reference value.

In the decisive year of building of a moderately prosperous society in all respects, the state proposed to build a “three-in-one” pattern of poverty alleviation and targeted poverty alleviation in accordance with local conditions in poverty-stricken counties. First of all, the problem of food and clothing in the countryside needs to be solved. Some crops of high economic value should be planted. Depending on the situation of different regions, the regression model is used to set up high-level indexes to score different plants, select suitable crops and then carry out planting planning according to simulated annealing algorithm to maximize the utilization of rural land. Secondly, to make the countryside rich, we need to develop our own characteristic products. At the same time, in order to avoid the problem of “depreciating land price by itself”, we need to make use of the current popular live broadcast to improve the popularity. According to the obtained data, different network platforms are scored and the greedy algorithm is used to select the optimal measurement standard and the optimal investment scheme is obtained through the classification processing of the optimal solution, so as to realize the application of mathematical modeling in the targeted poverty alleviation work.

In recent years, with the further development of deep learning and high-performance computing technologies, more intelligent algorithms have been applied to the actual production processes. The main research content of this paper is the prediction of element yield in the steel industry. First of all, considering that there is a lot of noise in the original data, it will affect the prediction accuracy, so this paper uses the "3-σ" principle and wavelet threshold denoising to preprocess the original data. Subsequently, this article starts from the effect of influencing factors on the yield, and uses a convolutional neural network (CNN) to predict the yield. By analyzing the prediction results, it is found that the prediction effect of CNN for some samples is not good. In order to optimize the model, this article starts from the two perspectives of the model and the algorithm. On the one hand, the LSTM neural network is introduced to take the historical data information of the yield into account, on the other hand, the CNN network and the LSTM are integrated through the Adaboost method. The advantages of the network are combined to establish an integrated learning algorithm of CNNLSTM-Adaboost, which further improves the prediction accuracy of the model. By simulating the model, a high-precision prediction result is obtained.

With the continuous development of the national big data strategy and the construction of "Digital China", cities have been given new connotations and requirements. This paper takes the real estate price changes in Hainan Province as the background, collects relevant information and data, and makes relevant analysis and reasonable predictions for the housing price changes in Hainan Province. This paper comprehensively considers the prediction defects of MEA_BP neural network, Elman neural network and Wavelet neural network, and uses a combination algorithm based on entropy method to comprehensively consider the results of all prediction models. On the one hand, it can improve the prediction accuracy of the model, and on the other hand, it can reduce the sensitivity of the prediction results to changes in a certain factor, making the prediction results more reliable. This paper establishes a multiple and effective new era smart city prediction model, which provides a reference for building a new city database supported by a new generation of information technology.

With the escalation of the trade war between China and the United States and the outbreak of the new coronavirus epidemic in early 2020, China’s economy has been seriously affected. Accurate prediction of future economic development is a necessary means to formulate economic development strategies. For this reason, taking Guangdong province, the province with the strongest economic vitality, as an example, this paper uses multiple regression method to calculate the correlation degree and influence weight of population number, number of enterprises and per capita disposable income with economic development, and obtains multiple regression linear equation. In addition, this paper also uses the NAR dynamic neural network model in machine learning algorithms to predict the future trends of the three factors and economic aggregates, and analyzes the feedback results of network training errors, autocorrelation values, and partial correlation values. Compared with the multiple regression method, it is found that the final results of the two are very similar, with small errors and high correlation.

Hex Chess attracts more and more people because of the simple rules, but also because of the simple rules, it brings complex algorithms. The simple Monte Carlo tree search process is often very slow due to the large amount of computation. In the double game, the Monte Carlo algorithm cannot converge to the best decision strategy of the double game. This paper proposes an improved algorithm of Monte Carlo tree search combined with MTD(f) algorithm, so that the search results are not distorted by the randomness of Monte Carlo algorithm. In order to further improve the computational efficiency of the search algorithm in the two-player game, the TCL and the connected domain strategy are adopted. Compared with the improved algorithm, not only shortens the time of Monte Carlo tree search, but also improves the performance.

At present, many countries in the world launch a large number of satellites into space and form satellite systems with different functions. Among them, a large number of satellites have similar functions, so many functions between satellite systems can actually be converted to each other. If the functions of a satellite system can be fully developed on the premise of its most advantageous functions, it can theoretically replace other systems in a comprehensive manner. As of now, Beidou satellite system is the most promising satellite system in China, and its positioning capability has reached the world's leading level. Advanced positioning functions and newly developed communication functions can be developed. In order to verify the feasibility of using target monitoring scheme, we design the wood and foam instead of the target equivalent wind tunnel simulation test, proved the feasibility of the plan and determine the time interval between the effective range and collecting data, and then designed a data "screen - collection - review" system is operable and application value. Finally, Beidou system can develop fluid monitoring functions similar to remote sensing satellites.

With the development of multi-intelligence related technologies and distributed artificial intelligence, people have gradually applied drones to maritime search and rescue work. Maritime search and rescue requires high efficiency and short search time, which is difficult, and how to use UAVs for efficient and rapid maritime search and rescue has become a major concern. In this paper, based on the extended square search algorithm, the original algorithm is improved, and the probability function of target discovery is introduced to make the algorithm closer to the actual search and rescue situation. Finally, based on the MATLAB platform, this paper simulates the algorithm by writing a program, and studies the relationship between flight height, search pitch and search width overlap ratio and search time. The simulation results show that the flying height of the drone is inversely proportional to the search time, and the overlap ratio is proportional to the time. In the aspect of emergency data backhaul and communication, based on the original agreement of Beidou, the necessary transmission agreement keywords are added to facilitate the subsequent processing of the data.

In the study of prosthetic control technology, researchers usually use decoded electromyographic signals (EMG) to obtain amputees' motion intentions. Predicting and accurately classifying the intent of human gestures can be used not only for rehabilitation and entertainment robots, but also for artificial intelligence robots. It enables us to develop new, more natural methods of human-computer interaction. In this paper, we propose an advanced gesture recognition model. In this model, myoelectric signal of forearm measured by commercial sensor myo is used as input. To extract and classify these gestures, a long-term and short-term memory neural network is applied and compared with other deep learning algorithms. The experimental results show that the model proposed in this paper can recognize the same four gestures as Myo Armband's proprietary recognition system, with an average recognition accuracy of 97.75%, which has higher recognition accuracy than the comparison algorithm. The proposed deep learning model is robust, which means that it can recognize a user's gestures even if there is no user's data in the training data set.

With the outbreak of COVID-19 at the end of 2019, under the requirement of in-depth study and implementation of the overall national security concept, people's health level has become the focus of people's attention, and it is also the most basic and fundamental important indicator to reflect people's livelihood. Taking Shenzhen, a city with strong comprehensive economic level, as an example, this paper uses data processing to select six major influencing factors, such as medical treatment and environment, and uses the method of regression and fitting crossover analysis to establish the fitting curve between factors and people's health level for prediction, and obtains the regression equation. On this basis, T-S Fuzzy Neural Network (T-S FNN) is used to divide the evaluation grade of regression model, make an effective evaluation of multiple factors of people's physical health level, establish a comprehensive prediction evaluation model, and obtain the gradient grade of factors affecting people's physical health correlation and their own direct factors.

With the improvement of people's living standard, the general life of worry-free, people's thinking has also improved, people's awareness of electricity saving is becoming stronger and stronger. When buying electrical appliances, people gradually tend to buy energy-saving appliances, and the consultation on energy-saving appliances is also very high. According to preliminary statistics, the proportion of customers who purchase energy-saving appliances reaches 85%-90%. Therefore, it is necessary to design a robot that can use clean energy and control home appliances. This design is a robot that can independently search for light and charge and control household appliances. On the one hand, it can make use of the solar energy collected to feed itself to the robot. On the other hand, it can also be used to charge low-power household appliances (such as electric toothbrush, electric razor, charging lamp and mobile phone). In addition, if the owner forgets to turn off the appliances after leaving the house, the robot will automatically turn them off to save energy. In this paper, C# is used to make an upper position, which is used to simulate the photosensitive travel of the car in the room. It also uses the data of current, voltage, power and so on of home appliances to judge the state of home appliances through terminal analysis and detection, and implements control.

In recent years, intelligent fault diagnosis technology with deep learning algorithms has been widely used in industry, and they have achieved gratifying results. Most of these methods require large amount of training data. However, in actual industrial systems, it is difficult to obtain enough and balanced sample data, which pose challenges in fault identification and classification. In order to solve the problems, this paper proposes a data generation strategy based on Wasserstein generative adversarial network and convolutional neural network (WG-CNN), which uses generator and discriminator to conduct confrontation training, expands a small sample set into a high-quality dataset, and uses one-dimensional convolutional neural network (1D-CNN) to learn sample characteristics and classify different fault types. Experimental results over the standard Case Western Reserve University (CWRU) bearing fault diagnosis benchmark dataset showed that the proposed method has obvious and satisfactory fault diagnosis effect with 100% classification accuracy for few-shot learning. In different noise environments, this method also has excellent performance.

Based on the relevant data of the talent market in A city, this paper analyzes the local talent demand situation. Firstly, all industries are divided into seven categories, and the local talent demand situation is analyzed from the three aspects of employment expectation, expected occupation and expected education background. The ratio of report ratio, industry employment demand and high academic qualifications are selected. The data is processed using wavelet threshold denoising and a NAR neural network model is established for prediction. In terms of employment, the relevant literature was consulted, and the employment rate, average salary and employment satisfaction data of Chinese students were collected, and the employment situation was predicted by combining the GM (1, 1) prediction model. Finally, starting from the number of applicants, the number of admissions, academic qualifications and the seven fast-growing industries in the city A, the SOM neural network model is established to compare the above-mentioned indicators of the city A with other cities in China, matching the same characteristics as the city A. Building an “intelligent forecasting and decision-making evaluation system” to help government departments and related institutions predict future urban development related information.

Talent is an important resource in the 21st century. For enterprises, cities and countries, the number of them is undoubtedly the key to increasing competitiveness. Cross-analysis was carried out on China’s first-line and new-tier cities, and the artificial intelligence industry was developed horizontally and vertically. Taking Xi’an, the bridgehead in the “Belt and Road” as an example, select Xi’an “GDP, social fixed investment, undergraduate and above talent demand ratio, undergraduate and above talent supply and demand ratio and the proportion of talent demand in various industries” indicators, using improved PSO algorithm to optimize wavelet neural network. The demand for talent in the seven major industries in Xi’an is predicted, and the employment situation of the IT industry is analyzed. Finally, based on the forecast results, we put forward decision-making suggestions for the future development of Xi’an.

Self-organizing feature map (SOM) neural network is a kind of competitive neural network with unsupervised learning. It has the strong abilities of self-organization and self-learning. However, the classification accuracy of SOM neural network may decrease when the features of tested object are not obvious. In this paper, the particle swarm optimization (PSO) algorithm is used to optimize the weight values of SOM network. Three indexes, i.e., intra-class density, standard deviation and sample difference, are used to judge the weight value, which can improve the classification accuracy of the SOM network. PSO–SOM network is applied to the detection of series arc fault in electrical circuits and compared with conventional SOM network and learning vector quantization (LVQ) network. The detection accuracy of the PSO–SOM network is 95%, which is higher than conventional SOM network and LVQ network.

Aiming at covering the shortage of single source sensor imaging and improving the contrast between the target and the background in image, this paper proposes an adaptive weight fusion algorithm of infrared and visible image based on a High-frequency Domain Convolutional Neural Network (HDCNN). Firstly, the high and low frequency components of the original image are obtained by using the Daubechies wavelet transform, and then a high-frequency domain convolutional neural network which can detect the frequency information ratio of infrared and visible light in the high-frequency subband is trained. Secondly, the network is used to perform adaptive weight fusion for the high frequency components and regional energy is used for fusion of the low frequency components. Finally, the fusion image is obtained by inverse wavelet transform. A large number of experiments have proved that the algorithm in this paper has a greater improvement over similar comparison algorithms in objective evaluation metrics such as standard deviation, spatial frequency and average gradient. The algorithm enhances the contrast between the target and the background in the fusion image, and enriches the characteristic information of the target itself.

In recent years, with the advent of the third wave of artificial intelligence, a large number of new technologies have been applied in social software and information services, such as big data collection and analysis, pattern recognition and user behavior prediction. The subsequent impact of mobile new media on a new generation of young people and the entire society has become one of the most promising and research issues. This paper obtains relevant data through online and offline surveys and authoritative organizations, uses Cronbach alpha coefficient to verify the validity of the survey results. The Spearman Rank correlation coefficient is calculated to analyze the relationship between the grade level of college students, the degree of influence on mobile new media and the duration of using mobile new media. Based on BP neural network and Elman neural network, the two models are combined by the optimal weighting method to predict the arrival rate of mobile new media information, and the conclusion based on quantitative analysis of mathematical model is given. The results show that the arrival rate of new media information on different platforms is on the rise in the future.

Agent and multi-agent technology originates from the research of distributed artificial intelligence. The traditional control methods of multi-agent system are centralized and hierarchical from top to bottom, which violates the idea of the adaptability, flexibility and relative independence of agents, and has the fatal defect of over-dependence on the main controller. In this paper, an artificial fish swarm algorithm and particle swarm optimization algorithm are combined to find the optimal solution and solve the problem of multi-agent cooperative work and path planning. Simulation results show that the multi-agent cooperative evolution based on AFSA-PSO algorithm can adapt to the dynamic changes of new tasks, and has not only a good global convergence, but also a fast local convergence rate, simple operation and few parameters.

Ecological environment, which human beings depend on for survival, constantly provides us with various benefits. The exploitation and use of land by human beings will cause damage to the ecological environment. More seriously, the loss caused by environmental damage is often not concluded in the cost. Therefore, it is necessary to establish a complete environmental cost assessment model. By analyzing the value of ecosystem service and combining with previous studies, we proposed a new model—“DIP- PRSC model”—for evaluating the environmental costs of land projects. This model creatively divides environmental costs into fixed costs and floating costs. We use 16 indicators to measure the cost of each part of the model. We applied the model to the construction of the Rondônia highway in the Brazilian to verify the validity of the model, using NAR neural network to predict the cost of each part of the DIP model, and put the predicted cost into the DIP model to calculate the total cost. The results show that the environmental cost of the land project will continue to rise in the next eight years.