Due to the complex and variable structure of fundus vessels， and the low contrast between the fundus vessel and the background， there are huge difficulties in segmentation of fundus vessels， especially small fundus vessels. U-Net based on deep fully convolutional neural network can effectively extract the global and local information of fundus vessel images，but its output is grayscale image binarized by a hard threshold， which will cause the loss of vessel area， too thin vessel and other problems. To solve these problems， U-Net and Pulse Coupled Neural Network （PCNN） were combined to give play to their respective advantages and design a fundus vessel segmentation method. First， the iterative U-Net model was used to highlight the vessels， the fusion results of the features extracted by the U-Net model and the original image were input again into the improved U-Net model to enhance the vessel image. Then， the U-Net output result was viewed as a gray image， and the PCNN with adaptive threshold was utilized to perform accurate vessel segmentation. The experimental results show that the AUC （Area Under the Curve） of the proposed method was 0.979 6，0.980 9 and 0.982 7 on the DRVIE， STARE and CHASE_DB1 datasets， respectively. The method can extract more vessel details， and has strong generalization ability and good application prospects.

In order to solve the problems of small targets， large noises， and many types in the logo recognition for vehicles on traffic road， a method combining a target detection algorithm based on deep learning and a template matching algorithm based on morphology was proposed， and a recognition system with high accuracy and capable of dealing with new types of vehicle logo was designed. First， K-Means++ was used to re-cluster the anchor box values and residual network was introduced into YOLOv4 for one-step positioning of the vehicle logo. Secondly， the binary vehicle logo template library was built by preprocessing and segmenting standard vehicle logo images. Then， the positioned vehicle logo was preprocessed by MSRCR （Multi-Scale Retinex with Color Restoration）， OTSU binarization， etc. Finally， the Hamming distance was calculated between the processed vehicle logo and the standard vehicle logo in the template library and the best match was found. In the vehicle logo detection experiment， the improved YOLOv4 detection achieves the higher accuracy of 99.04% compared to the original YOLOv4， two-stage positioning method of vehicle logo based on license plate position and the vehicle logo positioning method based on radiator grid background； its speed is slightly lower than that of the original YOLOv4， higher than those of the other two， reaching 50.62 fps （frames per second）. In the vehicle logo recognition experiment， the recognition accuracy based on morphological template matching is higher compared to traditional Histogram Of Oriented Gradients （HOG）， Local Binary Pattern （LBP） and convolutional neural network， reaching 91.04%. Experimental results show that the vehicle logo detection algorithm based on deep learning has higher accuracy and faster speed. The morphological template matching method can maintain a high recognition accuracy under the conditions of light change and noise pollution.

Aiming at the problem that the existing differential privacy high-dimensional data publishing methods are difficult to take into account both the complex attribute correlation between data and computational cost, a differential privacy high-dimensional data publishing method based on clustering analysis technology, namely PrivBC, was proposed. Firstly, the attribute clustering method was designed based on the K-means++, the maximum information coefficient was introduced to quantify the correlation between the attributes, and the data attributes with high correlation were clustered. Secondly, for each data subset obtained by the clustering, the correlation matrix was calculated to reduce the candidate space of attribute pairs, and the Bayesian network satisfying differential privacy was constructed. Finally, each attribute was sampled according to the Bayesian networks, and a new private dataset was synthesized for publishing. Compared with PrivBayes method, PrivBC method had the misclassification rate and running time reduced by 12.6% and 30.2% averagely and respectively. Experimental results show that the proposed method can significantly improve the computational efficiency with ensuring the data availability, and provides a new idea for the private publishing of high-dimensional big data.

The Self-Organized Migrating Algorithm (SOMA) only can solve the single task, and the "implicit parallelism" of SOMA is not fully exploited. Aiming at the shortcomings, a new Self-Organized Migrating Algorithm for Multi-task optimization with Information Filtering (SOMAMIF) was proposed to solve multiple tasks concurrently. Firstly, the multi-task uniform search space was constructed, and the subpopulations were set according to the number of tasks. Secondly, the current optimal fitness of each subpopulation was judged, and the information transfer need was generated when the evolution of a task stagnated in a successive generations. Thirdly, the useful information was chosen from the remaining subpopulations and the useless information was filtered according to a probability, so as to ensure the positive transfer and readjust the population structure at the same time. Finally, the time complexity and space complexity of SOMAMIF were analyzed. Experimental results show that, SOMAMIF converges rapidly to the global optimal solution 0 when solving multiple high-dimensional function problems simultaneously; compared with those of the original datasets, the average classification accuracies obtained on two datasets by SOMAMIF combing with the fractal technology to extract the key home returning constraints from college students with different census register increase by 0.348 66 percentage points and 0.598 57 percentage points respectively.

The advancement of network technologies such as bluetooth and WiFi has promoted the development of the Internet of Things(IoT). IoT facilitates people's lives, but there are also serious security issues in it. Without secure access control, illegal access of IoT may bring losses to users in many aspects. Traditional access control methods usually rely on a trusted central node, which is not suitable for an IoT environment with nodes distributed. The blockchain technology and smart contracts provide a more effective solution for access control in IoT applications. However, it is difficult to ensure the correctness of smart contracts used for access control in IoT applications by using general test methods. To solve this problem, a method was proposed to formally verify the correctness of smart contracts for access control by using model checking tool Verds. In the method, the state transition system was used to define the semantics of the Solidity smart contract, the Computation Tree Logic(CTL) formula was used to describe the properties to be verified, and the smart contract interaction and user behavior were modelled to form the input model of Verds and the properties to be verified. And then Verds was used to verify whether the properties to be verified are correct. The core of this method is the translation from a subset of Solidity to the input model of Verds. Experimental results on two smart contracts for access control of IoT source show that the proposed method can be used to verify some typical scenarios and expected properties of access control contracts, thereby improving the reliability of smart contracts.

Focusing on the task allocation problem in spatial crowdsourcing, with the consideration of the influence of the spatial crowdsourcing workers' service quality on the allocation results, a task allocation strategy with the quality evaluation of worker's service was proposed. Firstly, in each spatio-temporal environment, the evaluation element of spatial crowdsourcing workers was added to establish a multi-objective model that fully considers the service quality and distance cost of the workers. Secondly, the algorithm convergence speed was increased and the global optimization ability was improved by improving the initialization and coding strategy, position movement strategy and neighborhood search strategy of the discrete glowworm swarm optimization algorithm. Finally, the improved algorithm was used to solve the model. Experimental results on the simulated and real datasets show that, compared with other swarm intelligence algorithms, the proposed algorithm can improve the total score of task allocation by 2% to 25% on datasets with different scales. By considering the service quality of workers, the proposed algorithm can effectively improve the efficiency of task allocation and the final total score.

For the problems of encountering and overtaking in the process of in-and-out port of ships in the container terminals with two-way channel, a new ship dispatching optimization algorithm focusing on the service rules was proposed. Firstly, the realistic constraints of two-way channel and the safety regulations of port night sailing were considered at the same time. Then, a mixed integer programming model with the goal of minimizing the total waiting time of ships in the terminal was constructed to obtain the optimal in-and-out port sequence of ships. Finally, the branch-cut algorithm with embedded polymerization strategy was designed to solve the model. The numerical experimental results show that, the average relative deviation between the result of the branch-cut algorithm using embedded polymerization strategy and the lower bound is 2.59%. At the same time, compared with the objective function values obtained by the simulated annealing algorithm and quantum differential evolution algorithm, the objective function values obtained by the proposed branch-cut algorithm are reduced by 23.56% and 17.17% respectively, which verifies the effectiveness of the proposed algorithm. The influences of different safe time intervals of ship arriving the port and ship type proportions were compared in the sensitivity analysis of the scheme obtained by the proposed algorithm, providing the decision and support for ship dispatching optimization in container terminals with two-way channel.

Aiming at the problems of blurred image, large noise, and poor visual perception in the traditional image super-resolution reconstruction methods, a method of image super-resolution reconstruction based on hybrid deep convolutional network was proposed. Firstly, the low-resolution image was scaled down to the specified size in the up-sampling phase. Secondly, the initial features of the low-resolution image were extracted in the feature extraction phase. Thirdly, the extracted initial features were sent to the convolutional coding and decoding structure for image feature denoising. Finally, high-dimensional feature extraction and computation were performed on the reconstruction layer using the dilated convolution in order to reconstruct the high-resolution image, and the residual learning was used to quickly optimize the network in order to reduce the noise and make the reconstructed image have better definition and visual effect. Based on the Set14 dataset and scale of 4x, the proposed method was compared with Bicubic interpolation (Bicubic), Anchored neighborhood regression (A+), Super-Resolution Convolutional Neural Network (SRCNN), Very Deep Super-Resolution network (VDSR), Restoration Encoder-Decoder Network (REDNet). In the super-resolution experiments, compared with the above methods, the proposed method has the Peak Signal-to-Noise Ratio (PSNR) increased by 2.73 dB,1.41 dB,1.24 dB,0.72 dB and 1.15 dB respectively, and the Structural SIMilarity (SSIM) improved by 0.067 3,0.020 9,0.019 7,0.002 6 and 0.004 6 respectively. The experimental results show that the hybrid deep convolutional network can effectively perform super-resolution reconstruction of the image.

Aiming at the problems of low accuracy of driving command prediction, bulky model structure and a large amount of information redundancy in existing end-to-end autonomous driving methods, a new end-to-end autonomous driving model based on deep visual attention neural network was proposed. In order to effectively extract features of autonomous driving scenes, a deep visual attention neural network, which is composed of the convolutional neural network, the visual attention layer and the long short-term memory network, was proposed by introducing a visual attention mechanism into the end-to-end autonomous driving model. The proposed model was able to effectively extract spatial and temporal features of driving scene images, focus on important information and reduce information redundancy for realizing the end-to-end autonomous driving that predicts driving commands from sequential images input by front-facing camera. The data from a simulated driving environment were used for training and testing. The root mean square errors of the proposed model for prediction of the steering angle in four scenes including country road, highway, tunnel and mountain road are 0.009 14, 0.009 48, 0.002 89 and 0.010 78 respectively, which are all lower than the results of the method proposed by NVIDIA and the method based on the deep cascaded neural network. Moreover, the proposed model has fewer network layers compared with the networks without the visual attention mechanism.

To solve the problems of requiring a large number of samples, not associating with time information, and not using global navigation information in motion planning for autonomous driving based on machine learning, a motion planning method for autonomous driving with directional navigation based on deep spatio-temporal Q-network was proposed. Firstly, in order to extract the spatial features in images and the temporal information between continuous frames for autonomous driving, a new deep spatio-temporal Q-network was proposed based on the original deep Q-network and combined with the long short-term memory network. Then, to make full use of the global navigation information of autonomous driving, directional navigation was realized by adding the guide signal into the images for extracting environment information. Finally, based on the proposed deep spatio-temporal Q-network, a learning strategy oriented to autonomous driving motion planning model was designed to achieve the end-to-end motion planning, where the data of steering wheel angle, accelerator and brake were predicted from the input sequential images. The experimental results of training and testing results in the driving simulator named Carla show that in the four test roads, the average deviation of this algorithm is less than 0.7 m, and the stability performance of this algorithm is better than that of four comparison algorithms. It is proved that the proposed method has better learning performance, stability performance and real-time performance to realize the motion planning for autonomous driving with global navigation route.

To handle the issues of lacking paired training samples and inconsistency between frames in translation from virtual to real driving scenes, a video translation model based on Generative Adversarial Networks was proposed in this paper. In order to solve the problem of lacking data samples, the model adopted a “dual networks” architecture, where the semantic segmentation scene was used as an intermediate transition to build front-part and back-part networks, respectively. In the front-part network, a convolution network and a deconvolution network were adopted, and the optical flow network was also used to extract the dynamic information between frames to implement continuous video translation from virtual to semantic segmentation scenes. In the back-part network, a conditional generative adversarial network was used in which a generator, an image discriminator and a video discriminator were designed and combined with the optical flow network to implement continuous video translation from semantic segmentation to real scenes. Data collected from an autonomous driving simulator and a public data set were used for training and testing. Virtual to real scene translation can be achieved in a variety of driving scenarios, and the translation effect is significantly better than the comparative algorithms. Experimental results show that the proposed model can handle the problems of the discontinuity between frames and the ambiguity for moving obstacles to obtain more continuous videos when applying in various driving scenarios.

For the parallel machine scheduling problem of minimizing the maximum completion time, an Improved Discrete Artificial Bee Colony algorithm (IDABC) was proposed by considering the processing efficiency of the machine and the delivery time of the product as well as introducing the mathematical model of the problem. Firstly, a uniformly distributed population and a generation strategy of the parameters to be optimized were achieved by adopting the population initialization strategy, resulting in the improvement of the convergence speed of population. Secondly, the mutation operator in the differential evolution algorithm and the idea of simulated annealing algorithm were used to improve the local search strategy for the employed bee and the following bee, and the scout bee was improved by using the high-quality information of the optimal solution, resulting in the increasement of the population diversity and the avoidance of trapping into the local optimum. Finally, the proposed algorithm was applied in the parallel machine scheduling problem to analyze the performance and parameters of the algorithm. The experimental results on 15 examples show that compared with the Hybrid Discrete Artificial Bee Colony algorithm (HDABC), IDABC has the accuracy and stability improved by 4.1% and 26.9% respectively, and has better convergence, which indicates that IDABC can effectively solve the parallel machine scheduling problem in the actual scene.

Due to the characteristics of water itself and the absorption and scattering of light by suspended particles in the water, a series of problems, such as low Signal-to-Noise Ratio (SNR) and low resolution, exist in underwater images. Most of the traditional processing methods include image enhancement, restoration and reconstruction rely on degradation model and have ill-posed algorithm problem. In order to further improve the effects and efficiency of underwater image restoration algorithm, an improved image super-resolution reconstruction method based on deep convolutional neural network was proposed. An Improved Dense Block structure (IDB) was introduced into the network of the method, which can effectively solve the gradient disappearance problem of deep convolutional neural network and improve the training speed at the same time. The network was used to train the underwater images before and after the degradation by registration and obtained the mapping relation between the low-resolution image and the high-resolution image. The experimental results show that on a self-built underwater image training set, the underwater image reconstructed by the deep convolutional neural network with IDB has the Peak Signal-to-Noise Ratio (PSNR) and Structural SIMilarity (SSIM) improved by 0.38 dB and 0.013 respectively, compared with SRCNN (an image Super-Resolution method using Conventional Neural Network) and proposed method can effectively improve the reconstruction quality of underwater images.

Aiming at the danger and unsatisfactory effect of dense crowd evacuation in public places in emergency, a motion planning algorithm of robots for crowd evacuation based on Deep Q-Network (DQN) was proposed. Firstly, a human-robot social force model was constructed by adding human-robot interaction to the original social force model, so that the motion state of crowd was able to be influenced by the robot force on pedestrians. Then, a motion planning algorithm of robot was designed based on DQN. The images of the original pedestrian motion state were input into the network and the robot motion behavior was output. In this process, the designed reward function was fed back to the network to enable the robot to autonomously learn from the closed-loop process of "environment-behavior-reward". Finally, the robot was able to learn the optimal motion strategies at different initial positions to maximize the total number of people evacuated after many iterations. The proposed algorithm was trained and evaluated in the simulated environment. Experimental results show that the proposed algorithm based on DQN increases the evacuation efficiency by 16.41%, 10.69% and 21.76% respectively at three different initial positions compared with the crowd evacuation algorithm without robot, which proves that the algorithm can significantly increase the number of people evacuated per unit time with flexibility and effectiveness.

To address the problems that rule-based motion planning algorithms under constraints need pre-definition of rules and temporal features are not considered in deep learning-based methods, a motion planning model based on deep cascading neural networks was proposed. In this model, the two classical deep learning models, Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) network, were combined to build a novel cascaded neural network, the spatial and temporal features of the input images were extracted respectively, and the nonlinear relationship between the input sequential images and the output motion parameters were fit to achieve the end-to-end planning from the input sequential images to the output motion parameters. In experiments, the data of simulated environment were used for training and testing. Results show that the Root Mean Squared Error (RMSE) of the proposed model in four scenes including country road, freeway, tunnel and mountain road is less than 0.017, and the stability of the prediction results of the proposed model is better than that of the algorithm without using cascading neural network by an order of magnitude. Experimental results show that the proposed model can effectively learn human driving behaviors, eliminate the effect of cumulative errors and adapt to different scenes of a variety of road conditions with good robustness.

To deal with the difficulty and low performance of emergency crowd evacuation in public spaces, a crowd evacuation method using robots based on the social force model was proposed. A new human-robot social force model based on the original social force model was first developed, where the human-robot interaction from robots to pedestrians was added to the original social force model. And then, a new method using robot based on the human-robot social force model was presented to evacuate the crowd. After joining the crowd evacuation scenes, the robots can influence the motion of the surrounding pedestrians and reduce the pressure among the pedestrians by moving in the crowd, thus increasing the crowd motion speed and improving the efficiency of crowd evacuation. Two classical scenarios, including a group of crowd escaping from a closed environment and two groups of crowd moving to each other by crossing, were designed and simulated to test the proposed method, and the crowd evacuation method without robots was used for comparison. The experimental results demonstrate that the proposed method based on human-robot social force model can obviously increase the crowd motion speed and improve the efficiency of crowd evacuation.

To deal with the crowd supervision issue of low accuracy and poor real-time performance in public places, an abnormal crowd behavior detection approach based on motion saliency map was proposed. Firstly, the Lucas-Kanade method was used to calculate the optical flow field of the sparse feature points, then the movement direction, velocity and acceleration of feature points were computed after filtering the optical flow field both in time and space. In order to precisely describe the crowd behavior, the amplitude of velocity, the direction change, and the amplitude of acceleration were mapped to three image channels corresponding to R, G, and B, respectively, and the motion saliency map for describing the characteristics of crowd movement was fused by this way. Finally, a convolution neural network model was designed and trained for the motion saliency map of crowd movement, and the trained model was used to detect abnormal crowd behaviors. The experimental results show that the proposed approach can effectively detect abnormal crowd behaviors in real time, and the detection rate in the datasets of UMN and PETS2009 are more than 97.9%.

With the development of bioinformatics, gene expression microarray and image recognition, classification on high-dimensional and small-sample-size data has become a challenging task in data ming, machine learning and pattern recognition as well. High-dimensional and small-sample-size data may cause the problem of "curse of dimensionality" and overfitting. Feature selection can prevent the "curse of dimensionality" effectively and promote the generalization ability of classification mode, and thus become a hot research topic. Accordingly, some recent development of world-wide research on feature selection in high-dimensional and small-sample-size classification was briefly reviewed. Firstly, the nature of high-dimensional and small-sample feature selection was analyzed. Secondly, according to their essential difference, feature selection algorithms for high-dimensional and small-sample-size classification were divided into four categories and compared to summarize their advantages and disadvantages. Finally, challenges and prospects for future trends of feature selection in high-dimensional small-sample-size data were proposed.

Aiming at the problem of energy hole in the nodes non-uniform deployment network model based on the ring in Wireless Sensor Network (WSN), a Ring-based Clustering Algorithm for Nodes Non-uniform Deployment (RCANND) was proposed. The number of the optimal cluster heads in each ring was calculated by minimizing the energy consumption of each ring in the nodes non-uniform deployment network model. The cluster head selectivity was calculated by using the residual energy of the nodes, the distance from the base station, and the average distance from the neighbor nodes. The cluster head rotation was carried out with the cluster head selection sequence in cluster, and the number of cluster formation phases was reduced to improve the efficiency of network energy utilization. The proposed algorithm was tested in the simulation experiments, the experimental results show that, the average energy consumption fluctuation of nodes under the same radius but different nodes deployment models is very small. The average energy consumption fluctuation of nodes under the same nodes deployment model but different radiuses is not obvious. The network lifetime was defined as the survivability of 50% network nodes. In the case of non-uniform deployment of nodes, the network lifetime of the proposed algorithm is higher than that of Unequal Hybrid Energy Efficient Distributed algorithm (UHEED) by about 18.1% while it is also higher than that of Rotated Unequal Hybrid Energy Efficient Distributed algorithm (RUHEED) by about 11.5%. In the case of uniform deployment of nodes, the network lifetime of the proposed algorithm is higher than that of sub-Ring-based Energy-efficient Clustering Routing for WSN (RECR) by about 6.4%. The proposed algorithm can effectively balance the energy consumption under different nodes deployment models and prolong the network lifetime.

To deal with the problem of dynamic path planning for autonomous driving with avoidance of obstacles, a real-time dynamic path planning approach was proposed to avoid obstacles in real-time under the condition of knowing initial vehicle position, speed, orientation and the obstacle positions. Firstly, a base frame of the road was constructed using the continuity of the second derivative for cubic spline curves combined with the information of the road edges and lanes. Secondly, the s-q coordinate system was established using the position and orientation of the vehicle and the curvature of the road. Then a set of smooth curves from the current position to the destination were generated as the path candidates in the s-q coordinate system. Finally, considering the factors of safety, smoothness and continuity, a novel cost function was designed, and the optimal path was selected by minimizing the cost function. Various simulative roads were designed to test the proposed method in the experiments. The experimental results show that the proposed approach has the ability of planning a safe and smooth path for avoiding the obstacles on both single-lane roads and multi-lane roads with good real-time performance.

The existing image segmentation algorithms based on Snake model generally have the disadvantages of poor noise robustness, limited application range, easy leakage of weak edge and difficult to converge to small and deep concave boundary of contour curve. In order to solve the problems, a novel image segmentation algorithm based on Snake model was proposed. Firstly, the Laplacian operator with isotropic smoothness was replaced by the new chosen diffusion term. Secondly, the p-Laplacian functional was introduced into the smooth energy term to strengthen the external force in the normal direction. Finally, the edge-preserving term was used to keep the external force field parallel to the edge direction, so as to prevent the weak edge from leaking and promote the contour curve to converge to the small and deep concave boundary. The experimental results show that, the proposed model not only overcomes the drawbacks of the existing image segmentation algorithms based on Snake model, possesses better segmentation effect, improves the anti-noise performance and corner positioning accuracy obviously, but also consumes less time. The proposed model is suitable for segmenting noise images, medical images, and natural images with many weak edges.

Intention is the main method of crowdsourcing service in Crowd Sensing (CS), in view of the existing methods in the process of service without fully considering the effects on CS which are from the number of participants and malicious competition, a kind of Incentive Mechanism based on Reverse Vickrey Auction model (RVA-IM) method was proposed. Firstly, incentive mechanisms of crowdsourcing were studied in this paper, in combination with reverse auction and Vickrey auction, a reverse auction model oriented to task covering was built. Secondly, the in-depth analysis and research on the key technical problems involved in the model were conducted, such as task covering, reverse auction selection and reward implementation. Finally, the effectiveness of RVA-IM method was analyzed in five ways:computational efficiency, individual rationality, budget-balance, truthfulness and honesty. The simulation results show that, compared with IMC-SS (Incentive Mechanism for Crowdsourcing in the Single-requester Single-bid (SS)-model) and MSensing (Myerson Sensing) method, RVA-IM method is more effective and feasible. It can solve the problem of malicious competition in the existing methods, and improves the average rate of service completion by 21%.

Aiming at the defects that the existing methods based on Parametric Active Contour Models (PACM) cannot accurately locate to corners, and discontinuous edges were easily affected by the surrounding irrelevant information, a new method for image segmentation based on PACM was proposed. In this method, the edge preserving term was first constructed, which was introduced to active contour model of image segmentation, and the tangent direction of Laplace diffusion term still persisted, and then two weight parameters were introduced to control tangential direction and normal direction so that the accuracy and efficiency for segmentation were improved. Experimental results show that the proposed model can detect weak edges and locate accurately corners, meanwhile converges to the depth of concave boundary and reduce the impact of independent information on edge discontinuities. Furthermore, it overcomes the edge leakage and is very good for protecting image details. Both the efficiency and accuracy of segmentation are significantly improved in contrast with the edge preserving gradient vector flow models, the normalized gradient vector flow models and their improved models.

Since fuzzy rough sets induced by fuzzy equivalence relations can not quite accurately reflect decision problems described by numerical attributes among fuzzy concept domain, a fuzzy rough set model based on neighborhood relation called NR-FRS was proposed. First of all, the definitions of the rough set model were presented. Based on properties of NR-FRS, a fuzzy neighborhood approximation space reasoning was carried out, and attribute dependency in characteristic subspace was also analyzed. Finally, feature selection algorithm based on NR-FRS was presented, and feature subsets was constructed next, which made fuzzy positive region greater than a specific threshold, thereby getting rid of redundant features and reserving attributes that have a strong capability in classification. Classification experiment was implemented on UCI standard data sets, which used Radial Basis Function (RBF) support vector machine as the classifier. The experimental results show that, compared with fast forward feature selection based on neighborhood rough set as well as Kernel Principal Component Analysis (KPCA), feature number of the subset obtained by NR-FRS model feature selection algorithm changes more smoothly and stably according to parameters. Meanwhile, average classification accuracy increases by 5.2% in the best case and varies stably according to parameters.

Concerning the premature convergence problem in Particle Swarm Optimization (PSO) algorithm, a Diversity Feedback and Control PSO (DFCPSO) algorithm was proposed. In the process of search, the algorithm dynamically adjusted the algorithm parameters according to the feedback information of diversity; as a result, the distribution of iterations in the diversity curve was improved. When the population diversity or the variance of the population's fitness dropped to the given thresholds, the proposed algorithm would let the particle swarm initialize based on the repulsion of the global best position and fly away from the gathering area efficiently to search again, hence the diversity was controlled in a reasonable range, which avoided premature convergence. The experimental results on several well-known benchmark functions show that DFCPSO has stronger global optimization ability in the complicated problems and multi-modal optimization when being compared with the existing Diversity-Controlled PSO (DCPSO).

An effective Poisson denoising model based on fractional derivative for images with Poisson noise was proposed to improve the denoising effect. The model inherited the advantages of total variation model to eliminate noise. Furthermore, due to the advantage of property of amplitude-frequency in fractional differentiation, it can protect "weak information" well in processing specifics of image and texture characteristics. The numerical experimental results demonstrate that the proposed method of fractional variation to eliminate noise is better than traditional integer variation and can protect the detail characteristics of image edges.

Current works on the Approximate Pattern Matching with Wildcards and Length constraints (APMWL) problem can only cope with replacement operation. This paper proposed an Edit Distance Matrix (EDM) method based on dynamic programming and the Approximate Pattern Matching with EDM (APM) algorithm. APM can handle all approximate operations including insertion, replacement and deletion. Moreover, this paper extended APM to the APM-OF algorithm with a strict constraint condition that each character can be used at most once for pattern matching in a sequence. The experiments verify that both APM and APM-OF have significant advantages on matching solutions against other peers. The average improvement rates of matching compared to SAIL-Approx are up to 8.34% and 12.37% respectively. It also demonstrates an advantage on approximate pattern mining that the number of approximate patterns mined by APM-OF is 2.07 times of that mined by OneoffMining.

Compressed Sensing (CS) theory can reconstruct original images from fewer measurements using the priors of the images sparse representation. The CS theory was applied into the single-image Super-Resolution (SR), and a new reconstruction algorithm based on two-step iterative shrinkage and Total Variation (TV) sparse representation was proposed. The proposed method does not need an existing training set but the single input low resolution image. A down-sampling low-pass filter was incorporated into measurement matrix to make the SR problem meet the restricted isometry property of CS theory, and the TV regularization method and a two-step iterative method with TV denoising operator were introduced to make an accurate estimate of the image's edge. The experimental results show that compared with the existing super-resolution techniques, the proposed algorithm has higher precision and better performance under different magnification level, the proposed method achieves significant improvement (about 4~6dB) in Peak Signal-to-Noise Ratio (PSNR), and the filter plays a decisive role in the reconstruction quality.

In order to solve the energy limitation problem on wireless sensor network nodes, this paper proposed a new adaptive update asynchronous MAC protocol — AU-MAC protocol. This protocol combined the sleep-work state switching mode, asynchronous mode with adaptive update to effectively extend the network life. AU-MAC protocol improved channel usage efficiency by making use of sender monitoring and receiver activating data transfer. And, it established a neighbor node information table and introduced adaptive updating mechanism, to reduce the free monitor. The functions of AU-MAC protocol had been estimated on NS2 network simulation platform. It shows that, AU-MAC protocol improves the energy efficiency at the basis of maintaining the same throughput and end-end transit delay.

According to the high complexity of extracting feature of radar signal using image processing method, a new method for extracting feature was proposed. Firstly, the time-frequency distribution was gained based on the adaptive Gaussian kernel time frequency analysis, then through analyzing the physical meaning of each element, one dimension vector could be found through a simple arithmetic instead of the complicated method through processing the time-frequency figure with image processing means, so the real-time requirement for sorting radar signal could be satisfied. The simulation results verify the efficiency of the proposed algorithm. Additionally, the accuracy can be kept at a high level while the Signal-to-Noise Ratio (SNR) is low.