Image grayscale is to downscale a 3-dimensional color image into a 1-dimensional grayscale image. Due to the act of downscaling, the information of a 3-dimensional matrix is represented by a 1-dimensional matrix, and there will inevitably be information loss, which makes it very important to maintain the detail contrast information of the original color image to the maximum extent. In this regard, this paper proposes a two-way contrast retention model and algorithm based on cosine similarity. This model consists of two local contrast retention sub-models based on cosine similarity, and the two sub-models achieve complementary functions, which can play a contrast retention dephasing effect on the regions with large contrast and regions with small contrast in the original color image respectively, so that the total model can play a two-way contrast retention effect. In addition, the model is solved using a parameter discrete search strategy to improve the real-time performance of the algorithm.
Accurate rectal staging has great significance for surgery. However, it is difficult for radiologists to diagnose because of the heavy workload and the shortage of experts. In this paper, we proposed a deep neural network to automatically identify the rectal cancer T-stage. The network combines 3D DenseNet with 3D attention proposal network, which can use the 3D image as training data and take full advantage of the latent relation of each slice. The APN can distinguish slight differences of each rectal cancer T-stage, which use the last convolution layer to create the attention region that can generate multi-scale images. Moreover, the APN can focus on the tumor region and use a rectangle box to identify the related region, which is approximately regarded as network interpretation. The experiment is conducted on MRI images of 254 patients with rectal cancer of T1, T2, T3 stage. In contrast experiment, our proposed method outperforms other methods, so it is indicated that our method is effective.
In this paper, we study the architectures of space-air-ground integration network (SAGIN) proposed by domestic scientific research institutes, and put forward an collaborative federal learning architecture suitable for SAGIN to solve the problems of insecurity and low timeliness caused by traffic backhaul. An anomaly traffic detection method is proposed based on the requirements and characteristics of SAGIN. The problem that it is difficult to manually label and extract features in the traffic of SAGIN is solved through the improvement of deep learning algorithm. The challenge of lack of professionals labeling training set is solved by studying the method of semi supervision. The problem of artificial feature engineering is solved by studying the end-to-end anomaly traffic detection algorithm. Finally, we design a simulation environment for the anomaly traffic detection in SAGIN, and verify the feasibility and advanced nature of the proposed methods.
With the continuous improvement of deep learning algorithms, it has achieved considerable results in the fields of machine vision and natural language processing. In the face of problems that cannot be solved by traditional methods or are not effective, deep learning can achieve more desirable results through its powerful feature learning and mastering the laws of the problem. The lane recognition based on a convolutional neural network is proposed to address the problems of lane crushing of self-driving intelligent vehicles and the accuracy of traffic marker recognition, respectively. The traffic marker recognition model based on YOLO5, by collecting data and completing model training, finally achieves the average lane crush of the self-driving car less than 1 time, and the traffic marker recognition rate reaches 98.5%.
When the software in embedded system runs, the current may change dynamically in wide ranges. It is difficult to accurately measure the energy consumption of software in embedded system. Some existing methods can provide the functions that measuring the currents of different regions by using different ranges through adjusting shunt resistor or amplification. However, these methods are hard to give the value of current accurately during range switching, and bring some measuring errors. To address this problem, this paper designs hardware and software schemes to measure the current of software in embedded system by using small, medium and large ranges simultaneously. Further, a two-stage calibration method based on machine learning is proposed. And a Measurement System for Energy consumption of software in Embedded system (MSee) is presented. The experimental results show that both the average relative errors of MSee in small, medium and large ranges and the average relative errors of MSee in transitional neighborhood of ranges are better than those existing methods of range switching.
Medical imaging detection technology can improve the efficiency and accuracy of doctors’ diagnosis, it has been widely used in medicine. Recently, object detection methods based on deep learning have achieved remarkable detection results in medical imaging detection. However, when these methods are specifically applied to skull fracture detection, it is difficult to satisfy the requirements and there are still many missed detection and misdetection problems. In order to solve above problems, this paper proposed an improved YOLOv5 method to detect skull fracture in the computed tomography (CT) images of the skull. The version of YOLOv5 used in this paper is YOLOv5s. Compared with other versions of YOLOv5, YOLOv5s has fewer parameters and is faster. Skull fracture detection is essentially tiny object detection. In the proposed approach, external attention and 3D feature fusion methods are introduced into YOLOv5. These methods improve YOLOv5’s ability to detect tiny objects. Experimental verifications are conducted on the dataset provided by Fujian medical university union hospital. Experimental results show that our approach outperforms the state-of-art object detectors with an average precision (AP) of 88.08
Due to the complex distribution of web data and frequent updates under the cloud service architecture, it is easy to drive data disasters and data accidents. This leads to many problems in web data mining, such as how to choose the best algorithm to process the corresponding type of data, and how to effectively reduce the dimensionality of the web data in the cloud service architecture and process multiple target information and multiple customer needs at the same time. In order to overcome these problems, we transform the multi-information goal and multi-user demand (constraint conditions) in web data mining into a constrained multi-objective optimization model, and solve it by a constrained particle swarm multi-objective optimization algorithm. In order to make it easier for the constrained multi-objective particle swarm algorithm to solve different types of problems to find an effective solution set closer to the real Pareto front, a new manifold learning algorithm based on the constrained multi-objective particle swarm algorithm is built and used to solve this problem. Experiments results demonstrate that this can improve the service efficiency of cloud computing.
The 0–1 multidimensional knapsack problem is an NP-hard combinatorial optimization problem, which is widely used in life. The existing heuristic algorithm suffers from the problem of high time consumption or low solution accuracy when faced with a large-scale and high-dimensional knapsack problem. This paper proposes an adaptive search algorithm with scatter and tabu strategy. The algorithm integrates adaptive scatter search and solution-based tabu search into the evolutionary framework. In the search process, adaptive reference set strategies, repeated flip mutations and the neighborhood structure that limits the step size change are used. In order to verify the performance of the algorithm, the algorithm was tested with 60 benchmarks, compared with other algorithms in the international standard library. Experimental results show that the proposed algorithm is competitive with other algorithms.
Blockchain technology has been initially used in the development of tourism. Based on the core characteristics of blockchain technology, summarizing its application status in the tourism industry, analysis found that this technology has made certain progress in smart tourism construction, encrypted currency payment and accommodation industry management. Combined with domestic and foreign literature studies, it is found that the main driving force for the application of tourism blockchain technology comes from the practical demand for innovation and development within the tourism industry; at the same time, it is also found that the main obstacles to the promotion of blockchain technology involve technology scalability, legal and regulatory and Safety hazards and other aspects. With the development of the sharing economy, the application of blockchain technology in tourism industry management, information management, corporate financial management, tourism sharing economy system construction, supply chain management, credit management and customer loyalty management will reshape the traditional tourism business model. Therefore, it is recommended to strengthen the training of tourism blockchain technology talents and innovative team building, introduce corresponding regulatory systems and laws and regulations, implement classified supervision on the application of tourism blockchain technology, and actively guide tourism companies to reshape their organizational structure and culture to enhance tourists Experience quality, conform to diversified tourism consumption needs, and actively respond to technological changes in the era of intelligent interconnection.
A large number of increasingly complex multi-objective optimization problems have emerged in scientific research and engineering practice, especially high-dimensional multi-objective problems, which has become a problem in the field of intelligent optimization. In order to solve the shortcomings of multi-objective particle swarm optimization in high-dimensional optimization, a new fitness allocation and multi-criteria mutation strategy for high-dimensional particle swarm evolution (FAMCHPSO) is proposed by combining fuzzy information theory and new mutation methods. The algorithm combines fuzzy information theory to abandon the disadvantages of the traditional fitness allocation method of multi-objective optimization algorithm, and proposes a new fitness allocation method, which increases the pressure of population selection, eliminates the influence of external uncertain factors on the algorithm and simplifies the algorithm process, making it suitable for solving high-dimensional multi-objective optimization problems. A new multi-criteria mutation strategy is introduced to effectively perturb the multi-objective particle algorithm, effectively avoiding the algorithm to fall into a local optimum. The FAMCHPSO algorithm is compared with three other representative multi-objective evolution algorithms on the DTLZ series test function set. The simulation results show that the FAMCHPSO algorithm has a significant performance advantage in terms of convergence, diversity, and robustness.
Blockchain origins from bitcoin which runs successfully since it was launched in 2009. The three most significant properties of bitcoin are decentralization, trusted feature and transparent. For those reasons, recently, the blockchain technology draws increasing attentions both from the academic circle and the Internet industry for it is able to not only build up cryptocurrency but also implement a trusted log system, which ensures all the logs are not tampering. This paper, firstly introduces the basic principles and applications of blockchain, such as hash function, hash pointer, digital signature. and then implements a blockchain-based trusted log storage and verification system in MES system.
Traditional three-dimensional map construction mainly includes measurement modeling, tilt photography and synchronous positioning and map creation technology developed in recent years. These techniques either spend time cycle is long, the high professional requirements, high precision of three-dimensional modeling, either in a short period of time need a lot of data processing, delay effect is poor, cannot achieve more intelligent three-dimensional map, map building based on object model technology can achieve rapid modeling, real-time building, computational complexity is small. In the object model-based method, the collected data is objectified and processed, and the artificial intelligence algorithm is used to realize the purpose of fast recognition, fast prototype building, fast prototype modification, fast display of real map scene, so that the three-dimensional map scene is more intelligent.
Skull fracture detection is to identify and locate the fracture regions in the medical images of the skull. In fact, skull fracture detection is the task of object detection. Since the skull fracture regions are generally tiny and easily confused with the normal bone suturae of the skull, skull fracture detection presents difficult challenges. Currently, object detection methods based on deep learning have achieved remarkable performance. However, the ability of these methods to detect and locate tiny objects is poor and it is difficult to achieve the required detection accuracy for skull fracture detection. In order to solve above problems, the paper proposed a novel anchor-free object detection method with unpaired keypoints for skull fracture detection. The proposed method is called centripetal and centrifugal net (CCNet). CCNet uses the top-left corner and the center of the bounding box as the keypoints. To generate the bounding boxes, two kinds of shifts are used. One kind of shift is centripetal shift and the other is the centrifugal shift. CCNet predicts the position and two kinds of shifts of the keypoints. Experimental verifications are conducted on the dataset provided by Fujian medical university union hospital. Experimental results show that CCNet outperforms the state-of-art anchor-free detectors with a mean average precision (MAP) of 89.9%.
The development of deep learning and artificial intelligence promotes the emergence of smart agriculture which is getting more and more attention. In view of the current characteristics of passion fruit pest identification in natural environmental factors such as background, angle and light, as well as the dynamics and stability of the pest itself and the large scale and different forms of the population, a deep convolutional network is used to establish a model for passion fruit pest identification under complex factors. On the basis of convolutional neural network, the PyTorch framework was used to train the pest identification model. The models with the lowest preservation loss and the highest mean average precision were compared for analysis and evaluation. The experimental results showed that this method could accurately identify five passion fruit pests, including Thrips, Scarab beetles, Red spiders, Spodoptera litura Fabricius, Epicauta ruficeps, with the mean average precision of 98.96%.
Garbage sorting policy has been implemented in more and more cities, but it is difficult and costly to implement garbage classification due to people’s insufficient cognition of garbage types. In view of the above problems, a WeChat Mini Program for rapid garbage identification and sorting and is proposed to facilitate users to query the classification of garbage. The system server adopts Spring Boot framework technology, encapsulates the transmission data into JSON type, obtains the garbage classification data set through distributed crawler technology, and uses the convolution neural network of TensorFlow platform to train the garbage sorting model to realize garbage image recognition. The WeChat Mini program also provides a variety of garbage search methods such as photo, voice and text, which is convenient for users to use, to improve residents’ awareness of garbage sorting and promote global ecological environment governance.
Image registration plays an important role in medical image fusion and surgical navigation. Iterative nearest point algorithm (INPA) is a high-precision image registration algorithm, but it also has the problems of huge computation cost and low efficiency. Therefore, we use the differential evolution algorithm to optimize the search process of the iterative nearest point algorithm, so that to improve its registration efficiency. Then a mutation operator selection algorithm based on fitness roughness and a slack selection strategy for differential evolution algorithm were proposed to meet the requirements of high precision and low delay of medical image registration. The registration experiments of chest CT images show that the proposed differential evolution algorithm can not only accelerate the speed of iterative nearest point algorithm, but also improve its registration accuracy.
The slicing technology can meet the diversified and personalized service needs of 5G networks, schedule virtual networks dynamically and improve the operation efficiency of the network. The dynamic resource allocation algorithm among slices is the key for the application of slicing technology, which has also become a research hotspot in recent years. However, the current dynamic resource allocation algorithm of 5G network slicing has many problems such as insufficient fairness, difficulty in ensuring user satisfaction and low resource utilization rate, etc. Under the circumstances of a wide variety of 5G services, it is difficult to meet the needs of user satisfaction under differentiated service conditions only from the perspective of the network. From the perspective of Quality of Experience (QoE), a QoE evaluation system for 5G network is constructed, and a QoE-based dynamic resource allocation algorithm of 5G network slicing on the basis of this system is built in this research. The results of the simulation experiments indicate that the algorithm has significant effect in improving QoE of 5G communities, and it also has good performance in improving the resource utilization rate in the communities.
Rectal cancer is a disease with a high incidence rate and mortality rate. Accurate T-staging played an important role in the selection of treatment options for rectal cancer patients. Convolutional Neural Networks (CNNs) are widely used in various fields of medical images and have achieved great results. However, the MRI images of rectal cancer patients are 3D, and it is difficult for the general 2DCNN to extract the complete spatial information of the images. In this paper, a novel 3D deep convolutional neural network is proposed to extract features from MRI images of rectal cancer patients and automatically perform T-staging of tumors, which is very important for developing subsequent treatment plans and prognosis. After preprocessing the data, we divided the MRI images of real rectal cancer patients into a training set and a test set in the ratio of 8:2, and input the training set into our ResNet-based modified 3DCNN architecture for training. Since the rectal cancer dataset is characterized by small and unbalanced samples like the vast majority of medical image datasets, we first augment the dataset and use a random under-sampling method to handle the balanced dataset. The high accuracy on the test set validates the efficiency of our study, shows that 3DCNN has very good results in extracting spatial feature information from MRI images, and demonstrates that our system can outperform other classifiers.
In reinforcement learning, when the state space is enormous or infinite, it is not feasible to find the exact value for each state in the memory. A common way to tackle this problem is to adopt linear value function approximation technique. In this paper, we review some commonly used linear value function approximation algorithms, which divided into two classes. The first class is the so-called on-policy algorithms that mainly include temporal difference and least-squares-based methods. The second is the off-policy algorithms that use two strategies to solve the exploration-exploitation problem. From the perspective of algorithm design, on-policy is simple, but it does not take both exploration and exploitation into consideration when collecting data, while the off-policy can well balance the relation between exploration and exploitation. In the end, we first present some experiments to demonstrate the usefulness of linear value function approximation technique in the continuous and high dimensional scenes, and then in-class and out-class experiments are respectively conducted to further illustrate the performance of different approximation algorithms.
In order to better extract image features and effectively remove noise features, we propose an image denoising model based on GRU (Gate Recurrent Unit) and Dense block. The GRU contacts the previous state through the current state, which can more effectively extract noise features from the picture. The Dense block can effectively improve the feature propagation efficiency, reduce the problem of vanishing gradient, and optimize the process of feature acquisition. The denoising performance can be effectively improved through the cascade of GRU and Dense block. Our extensive evaluations on two datasets demonstrate that the proposed model outperforms the state of the art methods under all different noise levels in terms of PNSR, and the visual effects achieved by the proposed model are also better than the competing methods.
