The task of image captioning is an important branch of image understanding. It requires not only the ability to correctly recognize the image content, but also the ability to generate grammatically and semantically correct sentences. The traditional encoder-decoder based model cannot make full use of image features and has only a single decoding method. In response to these problems, a multi-layer encoding and decoding model for image captioning based on attention mechanism named MLED was proposed. Firstly, Faster Region-based Convolutional Neural Network (Faster R-CNN) was used to extract image features. Then, Transformer was employed to extract three kinds of high-level features of the image. At the same time, the pyramid fusion method was used to effectively fuse the features. Finally, three Long Short-Term Memory (LSTM) Networks were constructed to decode the features of different layers hierarchically. In the decoding part, the soft attention mechanism was used to enable the model to pay attention to the important information required at the current step. The proposed model was tested on MSCOCO dataset and evaluated by BLEU, METEOR, ROUGE-L and CIDEr. Experimental results show that on the indicators BLEU-4, METEOR and CIDEr, the model is increased by 2.5 percentage points, 2.6 percentage points and 8.8 percentage points compared to the Recall what you see (Recall) model respectively, and is improved by 1.2 percentage points, 0.5 percentage points and 3.5 percentage points compared to the Hierarchical Attention-based Fusion (HAF) model respectively. The visualization of the generated description sentences show that the sentence generated by the proposed model can accurately reflect the image content.

Fast and reliable hand pose estimation has a wide application in the fields such as human-computer interaction. In order to deal with the influences to the hand pose estimation caused by the light intensity changes, self-occlusions and large pose variations, a deep network framework based on label distribution learning was proposed. In the network, the point cloud of the hand was used as the input data, which was normalized through the farthest point sampling and Oriented Bounding Box (OBB). Then, the PointNet++ was utilized to extract features from the hand point cloud data. To deal with the highly non-linear relationship between the point cloud and the hand joint points, the positions of the hand joint points were predicted by the label distribution learning network. Compared with the traditional depth map based approaches, the proposed method was able to effectively extract discriminative hand geometric features with low computation cost and high accuracy. A set of tests were conducted on the public MSRA dataset to verify the effectiveness of the proposed hand pose estimation network. Experimental results showed that the average error of the hand joints estimated by this network was 8.43 mm, the average processing time of a frame was 12.8 ms, and the error of pose estimation was reduced by 11.82% and 0.83% respectively compared with the 3D CNN and Hand PointNet.

The various shapes of ferrite magnetic tiles and the wide varieties of their surface defects are great challenges for computer vision based surface defect quality recognition. To address this problem, the deep learning technique was introduced to the magnetic tile surface quality recognition, and a surface defect detection system for magnetic tiles was proposed based on convolution neural networks. Firstly, the tile target was segmented from the collected image and was rotated in order to obtain the standard image. After that, the improved multiscale ResNet18 was used as the backbone network to design the recognition system. During the training process, a novel within-class mixup operation was designed to improve the generalization ability of the system on the samples. To close to the practical application scenes, a surface defect dataset was built with the consideration of illumination changes and posture differences. Experimental results on the self-built dataset indicate that the proposed system achieves recognition accuracy of 97.9%, and provides a feasible idea for the automatic recognition of magnetic tile surface defects.

Aiming at the problem that traditional single detection algorithm models have low detection rate and slow detection speed on different types of attacks in industrial control system, an intrusion detection model combining optimized Support Vector Machine (SVM) and K-means++algorithm was proposed. Firstly, the original dataset was preprocessed by Principal Component Analysis (PCA) to eliminate its correlation. Secondly, an adaptive mutation process was added to Particle Swarm Optimization (PSO) algorithm to avoid falling into local optimal solution during the training process. Thirdly, the PSO with Adaptive Mutation (AMPSO) algorithm was used to optimize the kernel function and penalty parameters of the SVM. Finally, a K-means algorithm improved by density center method was united with the optimized support vector machine to form the intrusion detection model, achieving anomaly detection of industrial control system. The experimental results show that the proposed method can significantly improve the detection speed and the detection rate of various attacks.

To solve the problems that group effect is neglected when selecting dictionary bases in sparse coding models, and distance between a features and a dictionary base can not be effectively measured by Euclidean distance, Non-negative Local Sparse Coding algorithm based on Elastic net and Histogram intersection (EH-NLSC) was proposed. Firstly, with elastic-net model introduced in the optimization function to remove the restriction on selected number of dictionary bases, multiple groups of correlation features were selected and redundant features were eliminated, improving the discriminability and effectiveness of the coding. Then, histogram intersection was introduced in the locality constraint of the coding, and the distance between the feature and the dictionary base was redefined to ensure that similar features share their local bases. Finally, multi-class linear Support Vector Machine (SVM) was adopted to realize image classification. The experimental results on four public datasets show that compared with LLC (Locality-constrained Linear Coding for image classification) and NENSC (Non-negative Elastic Net Sparse Coding), the classification accuracy of EH-NLSC is increased by 10 percentage points and 9 percentage points respectively on average, proving its effectiveness in image representation and classification.

Artificial identification of ischemic stroke lesion is time-consuming, laborious and easy be added subjective differences. To solve this problem, an automatic segmentation algorithm based on 3D deep residual network and cascade U-Net was proposed. Firstly, in order to efficiently utilize 3D contextual information of the image and the solve class imbalance issue, the patches were extracted from the stroke Magnetic Resonance Image (MRI) and put into network. Then, a segmentation model based on 3D deep residual network and cascade U-Net was used to extract features of the image patches, and the coarse segmentation result was obtained. Finally, the fine segmentation process was used to optimize the coarse segmentation result. The experiment results show that, on the dataset of Ischemic Stroke LEsion Segmentation (ISLES), for the proposed algorithm, the Dice similarity coefficient reached 0.81, the recall reached 0.81 and the precision reached 0.81, the distance coefficient Average Symmetric Surface Distance (ASSD) reached 1.32 and Hausdorff Distance (HD) reached 22.67. Compared with 3D U-Net algorithm, level set algorithm, Fuzzy C-Means (FCM) algorithm and Convolutional Neural Network (CNN) algorithm, the proposed algorithm has better segmentation performance.

Focused on that limitation of single-layer structure on image feature learning ability, a deep architecture based on sparse representation of image blocks was proposed, namely Multi-layer incorporating Locality and non-negativity Laplacian Sparse Coding method (MLLSC). Each image was divided uniformly into blocks and SIFT (Scale-Invariant Feature Transform) feature extraction on each image block was performed. In the sparse coding stage, locality and non-negativity were added in the Laplacian sparse coding optimization function, dictionary learning and sparse coding were conducted at the first and second levels, respectively. To remove redundant features, Principal Component Analysis (PCA) dimensionality reduction was performed before the second layer of sparse coding. And finally, multi-class linear SVM (Support Vector Machine) was adopted for image classification. The experimental results on four standard datasets show that MLLSC has efficient feature expression ability, and it can capture deeper feature information of images. Compared with the single-layer algorithms, the accuracy of the proposed algorithm is improved by 3% to 13%; compared with the multi-layer sparse coding algorithms, the accuracy of the proposed algorithm is improved by 1% to 2.3%. The effects of different parameters were illustrated, which fully demonstrate the effectiveness of the proposed algorithm in image classification.

Since the vehicles provided by current online car rental platforms are in the fixed price, there are some issues coming up such as unreasonable allocation of the vehicle resources, unreliable price that could not indicates the real market supply and demand timely, and generally low social welfare. Therefore, an auction based vehicle allocation and pricing mechanism for car rental was proposed. Firstly, a mathematical model and a social welfare maximization objective function were established by studying the model of online car rental issues. Secondly, based on the minimum cost and maximum flow algorithm, the optimal vehicle resource allocation algorithm was adopted among the rental vehicle allocation algorithms. Finally, in terms of the price calculation algorithms, a truthful VCG (Vickrey-Clarke-Groves) price algorithm was used to calculate the final price. As a result, compared with the traditional first-come-first-serving algorithms, the order success rate of the proposed scheme was increased by 20% to 30%, and the revenue was increased by about 30%. Theoretical analysis and experiment results show that the proposed mechanism has the advantages of optimizing vehicle allocation and flexible price strategy.

The difficulty in sketch-based image retrieval is the effective recognition of images with different scales, positions, rotations and deformations. In order to identify and retrieve images of different scales, positions and rotations more accurately, a Sketch-Based Image Retrieval method Using Local Geometry Moment Invariant (SBIRULGMI) was proposed. Firstly, the geometric characteristics of image were used to determine the coordinate system of image. Secondly, the geometry moment invariant of image blocks which were divided averagely based on the generated coordinate system was calculated to form a eigenvector. Then, the similarities between query sketch and images in database were calculated based on Euclidean distance. Finally, the retrieval results were obtained from the similarity ranking and optimized according to Ant Colony Optimization (ACO). Compared with Shape Context (SC), Edge Orientation Histogram (EOH), GAbor Local lIne-based Feature (GALIF) and MindFinder, the retrieval accuracy of the proposed method in image database of MPEG-7 shape1 part B was increased by 17 percentage points on average. The experimental results show that the proposed method not only has a better recognition effect on the images after translation, scaling and flipping transformation, but also has better robustness to a certain degree of rotation and deformation.

The performances and correctness of system are affected by the state transition real-time process of the cyber-physical system. In order to solve the problem, aiming at the state transition process of actor's output-event driven system, a new scheduling strategy of value evaluation for output-event of actor named Value Evaluation-Information Entropy and Quality of Data (VE-IE&QoD) was proposed. Firstly, the real-time performance of event was expressed through the super dense time model. The self-information of the output-event, the information entropy of the actor and the quality of data were defined as the function indexes of value evaluation. Then, the value evaluation mission was executed for the process of the actor in performing task and it was considered about suitably increasing the weighting coefficient for parametric equation. Finally, the discrete event models which contain the proposed VE-IE&QoD scheduling strategy, the traditional Earliest Deadline First (EDF) scheduling algorithm and Information Entropy ^* (IE ^*) scheduling strategy were built by Ptolemy Ⅱ platform. The operation situation of different algorithm models was analyzed, the change of value evaluation and execution time of different algorithm models were compared. The experimental results show that, the VE-IE&QoD scheduling strategy can reduce the system average execution time, improve the memory usage efficiency and task value evaluation. The proposed VE-IE&QoD scheduling strategy can improve the system performance and correctness to some extent.

Aiming at the universality of individual interaction and the heterogeneity of individual social influence in opinion spreading, an opinion formation model of social network was proposed on the basis of Hegselmann-Krause model. By introducing the concepts of intimacy between individuals, interpersonal similarity and interaction strength, the individual interactive set was extended, the influence weight was reasonably quantified, and more realistic view of interaction rule was built. Through a series of simulation experiments, the effects of main parameters in the model on opinion evolution were analyzed. The simulation results indicate that group views can converge to the same and form consensus under different confidence thresholds. And the larger the confidence threshold is, the shorter the convergence time is. When confidence threshold is 0.2, convergence time is only 10. Meanwhile, extending the interactive set and increasing the strength of interpersonal similarity will promote consensus formation. Besides, when the clustering coefficient and the average degree of scale-free network are higher, the group views are more likely to produce convergence effect. The results are helpful to understand the dynamic process of opinion formation, and can guide social managers to make decisions and analysis.

The alert correlation approach based causal relationship has the problems that it cannot be able to process massive alerts in time and the attack scenario graphs split. In order to solve the problem, a novel real-time alert correlation approach based on Attack Planning Graph (APG) was proposed. Firstly, the definition of APG and Attack Planning Tree (APT) were presented. The real-time alert correlation algorithm based on APG was proposed by creating APG model on basis of priori knowledge to reconstruct attack scenario. And then, the attack scenario was completed and the attack was predicted by applying alert inference mechanism. The experimental results show that, the proposed approach is effective in processing massive alerts and rebuilding attack scenarios with better performance in terms of real-time. The proposed approach can be applied to analyze intrusion attack intention and guide intrusion responses.

Priority Belief Propagation (priority-BP) algorithm cannot satisfy real-time requirement, and there is much room to improve its computational efficiency. As for its application in image inpainting, the main improvements of priority-BP algorithm focused on information transmission and tag searching. In information transmission step, sparse representation of image was wielded into the first iteration and the initial image information of target area was rapidly updated to provide more accurate prior information for the first iteration, which accelerated information transmission and improved the accuracy of tag trimming and information transmission. In tag searching step, the global searching was integrated with local searching instead of just only global searching so as to improve the construction efficiency of tag set. The improved algorithm was verified by examples. The results show that it has obvious advantage in inpainting images with large size; even with the small size of 120×126, it still improves the Peak Signal-to-Noise Ratio (PSNR) by 1.1 dB compared with original priority-BP algorithm, and reduces time consumption up to 1.2 seconds compared with original priority-BP algorithm. The experimental results indicate that the propsed algorithm can effectively improve the inpainting accuracy and efficiency.

In order to improve the efficiency of power transmission line inspection by Unmanned Aerial Vehicle (UAV), a new method was proposed for detecting broken transmission lines and defects of foreign body based on the perception of line structure. The transmission line image acquired by UAV was easily influenced by the background texture and light, the gradient operators of horizontal and vertical direction which can be used to detect the line width were used to extract line objects in the inspection image. The study on calculation of gestalt perception of similarity, continuity and colinearity connected the intermittent wires into continuous wires. Then the parallel wire groups were further determined through the calculation of parallel relationship between wires. In order to reduce the detection error rate, spacers and stockbridge dampers of wires were recognized based on a local contour feature. Finally, the width change and gray similarity of segmented conductor wire were calculated to detect the broken part of wire and foreign object defect. The experimental results show that the proposed method can detect broken wire strand and foreign object defect efficiently under complicated backgrounds from the transmission line of UAV images.

Focusing on the issue that the label kernel functions do not take the correlation between labels into consideration in the multi-label feature extraction method, two construction methods of new label kernel functions were proposed. In the first method, the multi-label data were transformed into single-label data, and thus the correlation between labels could be characterized by the label set; then a new label kernel function was defined from the perspective of loss function of single-label data. In the second method, mutual information was used to characterize the correlation between labels, and a new label kernel function was proposed from the perspective of mutual information. Experiments on three real-life data sets using two multi-label classifiers demonstrated that the best method of all measures was feature extraction method with label kernel function based on loss function and the performance of five evaluation measures on average increased by 10%; especially on the data set Yeast, the evaluation measure Coverage reached a decline of about 30%. Closely followed by feature extraction method with label kernel function based on mutual information and the performance of five evaluation measures on average increased by 5%. The theoretical analysis and simulation results show that the feature extraction methods based on new output kernel functions can effectively extract features, simplify learning process of multi-label classifiers and, moreover, improve the performance of multi-label classification.

The traditional secret sharing schemes based on interpolation polynomial require a heavy computational cost. When the data is large, the efficiency of operation is particularly low. Therefore, a new secret sharing scheme for protecting the security of large scale data was proposed. The proposed scheme used the data block's method and need the exclusive-OR (XOR) operation over GF(2) only. The theoretical analysis and experimental results show that, compared to the traditional secret sharing scheme based on interpolation polynomial, the new scheme is increased by 19.3% in the operational efficiency.

Based on open source softwares of Computer Haptics, visualizAtion and Interactive in 3D (CHAI 3D) and Open Graphic Library (OpenGL), a virtual surgical system was designed for reduction of maxillary fracture. The virtual simulation scenario was constructed with real patients' CT data. A geomagic force feedback device was used to manipulate the virtual 3D models and output haptic feedback. On the basis of the original single finger-proxy algorithm, a multi-proxy collision algorithm was proposed to solve the problem that the tools might stab into the virtual organs during the simulation. In the virtual surgical system, the operator could use the force feedback device to choose, move and rotate the virtual skull model to simulate the movement and placement in real operation. The proposed system can be used to train medical students and for preoperative planning of complicated surgeries.

Images of transmission tower acquired by Unmanned Aerial Vehicle (UAV) have high resolution and complex background, the traditional stitching algorithm using feature points can detect a large number of feature points from background which costs much time and affects the matching accuracy. For solving this problem, a new image mosaic algorithm with quick speed and strong robustness was proposed. To reduce the influence of the background, each image was first segmented into foreground and background based on a new implementation method of salient region detection. To improve the feature point extraction and reduce the computation complexity, transformation matrix was calculated and image registration was completed by ORB (Oriented Features from Accelerated Segment Test (FAST) and Rotated Binary Robust Independent Elementary Features (BRIEF)) feature. Finally, the image mosaic was realized with image fusion method based on multi-scale analysis. The experimental results indicate that the proposed algorithm can complete image mosaic precisely and quickly with satisfactory mosaic effect.

Conventional Particle Swarm Optimization (PSO) algorithm has disadvantage of premature convergence and is easily trapped in local optima. An improved PSO algorithm with adaptive task allocation was proposed to avoid those disadvantages. Adaptive task allocation was applied to particles according to their distribution status and fitness. All the particles were divided into exploration particles and exploitation particles, and carried out different tasks with global model and dynamic local model respectively. This strategy can make better trade-off between exploration and exploitation and enhance the diversity of particle. Dynamic neighborhood strategy broadened the search space and effectively inhibited the premature stagnation. Gaussian disturbance learning was applied to the stagnant elite particles to help them jump out from local optima region. The superior performance of the proposed algorithm in global search ability and solution accuracy was validated by optimizing six complicated composition test functions.

Most existing stereoscopic image quality assessment methods convert color images to gray scale images, which loses the color information, so it is not conducive for color stereopairs to make the right assessment. To solve this problem, a quality assessment method of color stereopairs was proposed. Firstly, the new algorithm used Principal Component Analysis (PCA) image fusion to deal with the reference image pairs and the distortion image pairs to generate 2D color images. Secondly, the low-frequency coefficients were extracted from the 2D images by color wavelet transform respectively. The information of low-frequency coefficients were expressed in quaternion form. In other words, hue component' local mean of low-frequency coefficients was regarded as real part of quaternion, and three primary color components were regarded as the imaginary parts of quaternion. Finally, singular value feature vectors were gained by quaternion singular value decomposition. Cosine angle, Bhattacharyya distance and chi-square distance based on singular value feature vectors were taken as image quality evaluation indexes respectively. The method was tested on the LIVE 3D Image Quality Database, which included both symmetric and asymmetric distorted 3D images published by university of Texas. The linear correlation coefficient and Spearman Rank Order Correlation Coefficient (SROCC) achieved 0.919 and 0.923 in symmetric database. The results have high accordance with the subjective evaluation and reach the expected values.

Aiming at not satisfactory defogging effect of the traditional defogging algorithm based on polarized characteristics in heavy fog, a new color space conversion algorithm using dark channel prior for polarization image dehazing was proposed. Compared with the traditional imaging technology, polarization imaging detection technology has remarkable advantages in the target detection and recognition of complex environment. Intensity, polarization degree and polarization angel information are usually used to describe target's polarization information for polarization images. In order to combine the polarization information and defogging model, a method of color space transformation was adopted. Firstly, the polarization information was converted into the components of the brightness, hue, saturation in Hue-Intensity-Saturation (HIS) color space and then the HIS color space was mapped to the Red-Green-Blue (RGB) space. Secondly, the dark channel prior principle was applied to get the dark channel image with the combination of the atmospheric scattering model in haze weather. Finally, the atmospheric transmission rate was elaborated by using softmatting algorithm based on sparse prior of the image. The experimental results show that, compared with the existing polarization defogging algorithm, many technical specifications of defogged images such as standard deviation, entropy, average gradient of the proposed algorithm have been greatly improved in very low visibility conditions. The proposed algorithm can effectively enhance the global contrast in heavy fog weather and improve the identification capability for the polarized images.

To solve the bottleneck on performance and low reliability of single core node, the cloud monitor system based on Peer-to-Peer (P2P) monitor network was designed and implemented. On the hardware deployment, the monitor nodes were capsulated in application containers and distributed on different racks to build P2P monitor network. By establishing distributed storage clusters with non-relational database, remote data access and backup were supported to improve system reliability. On the software implementation, it was designed hierarchically using combination method of push and pull to collect data, estimating trust degree of data, storing data in distributed nodes and managing hosts in cloud with threshold control and free host estimation strategies. Through the system tests, it was concluded that the system only accounted for 2.17% on average computing resource usage and its average responsive ratio on reading and writing requests per millisecond reached above 93%.The results indicate that the monitor system is superior on low resource consumption and high read/write efficiency.

Focusing on the issue that the communication path selected by random routing algorithm of the onion router (Tor) can not be controlled, thus leading to problems such as the abuse of anonymous techniques and the failure of tracing methods, a Tor anonymous circuit control method based on node failure was proposed. To effectively control the circuit, the fake TCP reset information was sent to mimic the node failure, so that the Tor client would not stop choosing nodes until it selected the controlled ones. The results of theoretic analysis of Tor network path selection algorithm and the real test in a private Tor network composed of 256 onion routers demonstrate the effectiveness of the proposed approach. Compared with traditional methods which deploy high bandwidth routers to attract users to select the controlled nodes, the proposed method can improve the probability of choosing controlled entry node from 4.8% to about 60%, when entry guard was generally enabled by Tor client by default. The results also show, as the length of a controlled path increases, the success rate of building path decreases. Therefore the proposed method is suitable for controlling short paths.

The Multidimensional Knapsack Problem (MKP) is a typical multi-constraint combinatorial problem. In order to solve this problem, a Modified Binary Cuckoo Search (MBCS) algorithm was proposed. Firstly, with the help of classical binary code transformer, the Binary Cuckoo Search (BCS) algorithm was built; Secondly, the virus evolution mechanism and virus infection operation were introduced into the BCS. Specifically, on one hand, it made the position of bird's nest have mutation mechanism, which could improve the diversity of the population; on another hand, the main groups that consisted of nest position transmitted information cross the vertical generations and guided the global search, while the virus groups transfered evolutionary information cross the same generation through virus infection and guided the local search. These improved the convergence speed and decreased the probability of falling into the local optimum. Thirdly, the hybrid repair strategy for infeasible solutions was designed according to the characteristics of the MKP. At last, comparison experiments among the MBCS algorithm, Quantum Genetic Algorithm (QGA), Binary Particle Swarm Optimization (BPSO) algorithm and BCS algorithm were given on 15 different problems from ELIB and OR_LIB database. The experimental results show that the computational error and standard deviation of MBCS are less than 1% and 170, respectively, which shows the MBCS algorithm can achieve better solutions with good accuracy and robustness than QGA, BPSO and BCS algorithm. It is an effective algorithm in solving NP-hard problems such as the MKP.

To improve the speed of image reconstruction based on fan-beam Filtered Back Projection (FBP), a new optimized fast reconstruction method was proposed for polar back-projection algorithm. According to the symmetry feature of trigonometric function, the preprocessing projection datum were back-projected on the polar coordinates at the same time. During the back-projection data coordinate transformation, the computation of bilinear interpolation could be reduced by using the symmetry of the pixel position parameters. The experimental result shows that, compared with the traditional convolution back-projection algorithm, the speed of reconstruction can be improved more than eight times by the proposed method without sacrificing image quality. The new method is also applicable to 3D cone-beam reconstruction, and can be extended to multilayer spiral three-dimensional reconstruction.

To solve the problem of the regulation of badminton dynamic stable equilibrium, the particle influence coefficient method of feather piece was put forward. The method combined badminton quality models and quality feather piece, bending camber degree, angle of attack, and other related factors. The feather piece of particle influence coefficient was obtained by adjusting the height centroid which satisfied badminton dynamic stability requirements got by striking tilt minimum square. Compared with the traditional badminton dynamic stabilization which must depend on the experience accumulated for a long time, the badminton particle influence coefficient method of feather piece that was put forward by this paper formed a theoretical system. And it had less time consumption, high efficiency, etc. The numerical results show that the proposed method is correct and effective.

Concerning that conventional Particle Swarm Optimization (PSO) is easy trapped in local optima and with low search efficiency in later stage, an improved PSO based on mean information and elitist mutation, named MEPSO, was proposed. Average information of swarm was introduced into MEPSO to improve the global search ability, and Time-Varying Acceleration Coefficient (TVAC) strategy was adopted to balance the local search and global search ability. In the latter stage of the iteration, the Cauchy mutation operation was applied to the global best particle to improve the global search ability and to further reduce the risk of trapping into local optimum. Contrast experiments on six benchmark functions were given. Compared with Basic PSO (BPSO), PSO with TVAC (PSO-TVAC), PSO with Time-Varying Inertia Weight factor (PSO-TVIW) and Hybrid PSO with Wavelet Mutation (HPSOWM), MEPSO achieved better mean value and standard variance with shorter optimization time and better reliability. The results show that MEPSO can better balance the ability of local search and global search, and can converge faster with higher accuracy and efficiency.

Images captured in hazy weather suffer from poor contrast and low visibility. This paper proposed a single image defogging algorithm to remove haze by combining with the characteristics of HSI color space. Firstly, the method converted original image from RGB color space to HSI color space. Then, based on the different affect to hue, saturation and intensity, a defogged model was established. Finally, the range of weight in saturation model was obtained by analyzing original images saturation, then the range of weight in intensity model was also estimated, and the original image was defogged. In comparison with other algorithms, the experimental results show that the running efficiency of the proposed method is doubled. And the proposed method effectively enhances clarity, so it is appropriate for single image defogging.

The traditional K-means algorithm has an increasing number of iterations, and often falls into local optimal solution and unstable clustering since the initial cluster centers are randomly selected. To solve these problems, an initial clustering centers selection algorithm based on Latent Dirichlet Allocation (LDA) model for the K-means algorithm was proposed. In this improved algorithm, the top-m most important topics in text corpora were first selected. Then, the text corpora was preliminarily clustered based on the m dimensions of topics. As a result, the m cluster centers could be got in the algorithm, which were used to further make clustering on all the dimensions of the text corpora. Theoretically, the center for each cluster can be determined based on the probability without randomly selecting them. The experiment demonstrates that the clustering results of the improved algorithm are more accurate with smaller number of iterations.

Complex Event Processing (CEP) is currently an advanced analytical technology which deals with high velocity event streams in a real-time way and primarily gets applied in Event Driven Architecture (EDA) system domain. It is helpful to realize intelligent business in many applications. For the sake of reporting its research status, the paper introduced the basic meaning and salient feature of CEP, and proposed a system architecture model composed of nine parts. Afterwards, the main constituents of the model were reviewed in terms of key technology and its formalization. In order to illustrate how to use CEP in the logistic Internet of things, an application framework with CEP infiltrating in it was also proposed here. It can be concluded that CEP has many merits and can play an important role in application fields. Finally, the shortcomings of this research domain were pointed out and future works were discussed. The paper systematically analyzed the CEP technology in terms of theory and practice so as to further develop CEP technology.