To address the problem of low recognition accuracy caused by vanishing gradient and overfitting in the cattle eye image feature extraction process, an improved DenseNet based cattle eye image feature extraction method was proposed. Firstly, the Scaled exponential Linear Unit (SeLU) activation function was used to prevent the vanishing gradient of the network. Secondly, the feature blocks of cattle eye images were randomly discarded by DropBlock, so as to prevent overfitting and strengthen the generalization ability of the network. Finally, the improved dense layers were superimposed to form an improved Dense convolutional Network (DenseNet). Feature information extraction recognition experiments were conducted on the self-built cattle eyes image dataset. Experimental results show that the recognition accuracy, precision and recall of the improved DenseNet are 97.47%, 98.11% and 97.90% respectively, and compared to the network without improvement, the above recognition accuracy rate, precision rate, recall rate are improved by 2.52 percentage points, 3.32 percentage points, 2.94 percentage points respectively. It can be seen that the improved network has higher precision and robustness.

In order to quickly calculate the optimal fitting position of the orthopedic plate on the surface of broken bone to reduce the repeated adjustment times of the orthopedic plate during the surgical operation, a rapid calculation method of orthopedic plate fit based on improved Iterative Closest Point (ICP) algorithm was proposed. Firstly, under the guidance of the doctor, the fitting area was selected on the surface of the broken bone, and the point cloud of the inner surface for the orthopedic plate was extracted by using the angle between the normal vectors of the surface points for the orthopedic plate. Then, the two groups of point cloud models were smoothed, and the grid sampling method was adopted to simplify the point cloud models, after these operations, the characteristic relationship between the point clouds was used for the initial registration. Finally, the boundary and internal feature key points of the inner surface point cloud model of the orthopedic plate were extracted, K-Dimensional Tree (KD-Tree) was used to search the adjacent points, so that the feature key points of the orthopedic plate and the selected area of the broken bone surface were accurately registered by ICP. Taking tibia as the example to carry out experiments, and the results show that the proposed method can improve the registration efficiency while maintaining relatively high registration degree compared with other registration algorithms proposed in recent years. The proposed algorithm can realize the rapid registration between different damage types of tibia and orthopedic plate, and it is universal to other damaged bones.

The carrier-based aircrafts on the carrier deck are dense and occluded, so that the carrier-based aircraft targets are difficult to detect, and the detection effect is easily affected by the lighting condition and target size. Therefore, an improved Faster R-CNN (Faster Region with Convolutional Neural Network) carrier-based aircraft target detection method was proposed. In this method, a loss function with a repulsion loss strategy was designed, and combined with multi-scale training, pictures collected under laboratory condition were used to train and test the deep convolutional neural network. Test experiments show that compared with the original Faster R-CNN detection model, the improved model has a better detection effect on occluded aircraft targets, the recall increased by 7 percentage points, and the precision increased by 6 percentage points. The experimental results show that the proposed improved method can automatically and comprehensively extract the characteristics of carrier-based aircraft targets, solve the detection problem of occluded carrier-based aircraft targets, has the detection accuracy and speed which can meet the actual needs, and has strong adaptability and high robustness under different lighting conditions and target sizes.

As most existing enterprise community discovery algorithms focus on homogenous market environment, without reflecting the participation of some enterprises in multiple supply chain operations, a core community representation model based on node mapping relationship, Map-Community, was proposed. By constructing two different role nodes and their different mapping relationships, the ownership community of a enterprise was determined. Based on this representation model, Node Mapping Algorithm (NMA) with approximately-linear time-space complexity was proposed. Firstly, filtering operation was used to obtain the biconnected core graph in the topology diagram of the supply chain network. Secondly, mapping degree was introduced to select the core enterprise nodes. Thirdly, local expansion was performed according to the mapping judgment rules. Finally, the local community structure was extended to the global network by backtracking and overlapping areas were discovered. In the LFR (Lancichinetti-Fortunato-Radicchi) network application experiment, NMA shows low sensitivity to threshold change and is superior to LFM (Local Fitness Maximization), COPRA (Community Overlap PRopagation Algorithm) and GCE (Greedy Clique Expansion) in terms of practicality. Simulation was carried out in the enterprise social network, and the meaning of distribution effect was summarized by the community division. The experimental results verify the feasibility of this algorithm for overlapping enterprise community discovery and its performance advantages in discovery quality.

Fuzzy rule-based models are widely used in many fields. The existing performance indexes for the models are mainly numeric, which ignore the characteristic of fuzzy sets in the models. Aiming at the problem, a new method of evaluating the performance of fuzzy rule-based models was proposed, to effectively evaluate the non-numeric (granular) nature of results formed by the fuzzy models. In this method, different from the commonly used numeric performance indexes (such as Mean Squared Error (MSE)), the characteristics of information granules were used to represent the quality of granular results output by the model and this proposed index was applied for the performance optimization of the fuzzy model. The performance of information granule was quantified by two basic indexes, coverage rate (of data) and specificity (of information granule itself), and the maximization of the output quality of granularity (expressed as the product of coverage rate and specificity) was realized with the use of particle swarm optimization. Moreover, the distribution of information granules formed through fuzzy clustering was optimized. The experimental results show the effectiveness of the proposed method on the performance evaluation of fuzzy rule-based models

Based on comparison possibility degree of Distributed Preference Relation (DPR) considering four pairwise relations of alternatives including superiority, inferiority, indifference and uncertainty, a multi-attribute decision making method with unknown attribute weights was proposed. Firstly, DPRs were transformed to score intervals by using grade scores. Based on the score intervals, the comparison possibility degree of DPR with consideration of uniform distribution was constructed. Secondly, the relationship between comparison possibility degree and the difference range of values of the proposed possibility degree and the possibility degree without considering probability distribution were theoretically analyzed and proven. At last, by using the proposed possibility degree, a multiple attribute decision making process considering unknown attribute weights was generated. Meanwhile, a high recognizable decision solution was created. The proposed method was applied to evaluate the strategic emerging industries of Wuhu, which verified the applicability and validity of the method.

An imbalance multi-class AdaBoost algorithm with active learning was proposed to improve the recognition accuracy of minority class in imbalance classification. Firstly, active learning was adopted to select the most informative samples for classifiers through multiple iterations of sampling. Secondly, a new sample selection strategy based on uncertainty of dynamic margin was proposed to tackle the problem of data imbalance in the multi-class case. Finally, the cost sensitive method was adopted to improve the multi-class AdaBoost algorithm: giving different class with different misclassification cost, adjusting sample weight update speed, and forcing weak learners to "concern" minority class. The experimental results on clinical TransThoracic Echocardiography (TTE) data set illustrate that, when compared with multi-class Support Vector Machine (SVM), the total recognition accuracy of heart disease increases by 5.9%, G-mean improves by 18.2%, the recognition accuracy of Valvular Heart Disease (VHD) improves by 0.8%, the recognition accuracy of Infective Endocarditis (IE) (minority class) improves by 12.7% and the recognition accuracy of Coronary Artery Disease (CAD) (minority class) improves by 79.73%; compared with SMOTE-Boost, the total recognition accuracy of heart disease increases by 6.11%, the G-mean improves by 0.64%, the recognition accuracy of VHD improves by 11.07%, the recognition accuracy of Congenital Heart Disease (CHD) improves by 3.67%. The experiment results on TTE data and 4 UCI data sets illustrate that when used in imbalanced multi-class classification, the proposed algorithm can improve the recognition accuracy of minority class effectively, and upgrade the overall classifier performance while guaranteeing the recognition accuracy of other classes not to be decreased dramatically.

Concerning the high time complexity in influence maximization under budget control, a fast influence maximization algorithm, namely BCIM, was proposed. Firstly, a new information dissemination model which propagates the initial nodes for many times was proposed. Secondly, the nodes with high influence ranking value were selected as candidate seeds, and the calculation of node's influence scope was decreased based on the short distance influence. Lastly, only one seed was selected at most in each set of candidate seeds by using the dynamic programming method. The experimental results show that, compared with Random (random algorithm), Greedy_MII (greedy algorithm based on the maximum influence increment) and Greedy_MICR (greedy algorithm based on the maximum of influence increment over cost ratio), the influence scope of BCIM is near to or a bit better than that of Greedy_MICR and Greedy_MII, but much worse than that of Random; the quality of seeds set of BCIM, Greedy_MICR and Greedy_MII is similar, but much better than that of Random; the running time of BCIM is several times of Random, while the running time of the both greedy algorithms are hundreds times of BCIM. In summary, BCIM algorithm can find a more effective seeds set in a short time.

The traditional road traffic accident forecast mainly uses the historical data, including the number and the loss of traffic accidents, to predict the future trend, however, the traditional method can not reflect the relationship between the traffic accident and real-time traffic characteristics, and it also can not prevent accidents effectively. In order to solve the problems above, a real-time traffic accident prediction method based on AdaBoost classifier was proposed. Firstly, the road traffic states were divided into normal conditions and dangerous conditions, and the real-time collection of traffic flow data was used as the characteristic variable to characterize the different states, so the real-time prediction problem could be converted to a classification problem. Secondly, the Probability Density Function (PDF) of traffic flow characteristics under the two conditions in different time scales were estimated by Parzen window nonparametric estimation method, and the estimated density function was analyzed by the separability criterion based on probability distribution, then the sample data with appropriate characteristic variable and time scale could be determined. Finally, the AdaBoost classifier was trained to classify different traffic conditions. The experimental results show that the correct ratio by using standard deviation of traffic flow characteristics to classify test samples is 7.9% higher than that by using average value. The former can reflect the differences of different traffic states better, and also get better classification results.

Focusing on the limitation of conventional static Functional Connectivity (FC) techniques in investigating the dynamic functional brain states, an effective method based on whole-brain Dynamic Functional Connectivity (DFC) was proposed to characterize the time-varying brain states. First, the Diffusion Tensor Imaging (DTI) data were used to construct individual whole-brain networks with high accuracy and the functional Magnetic Resonance Imaging (fMRI) data of motor-related task was projected to the corresponding DTI space to extract the fMRI signals of each node for each subject. Then, one kind of sliding time window approach was applied to calculate the time-varying whole-brain functional connectivity strength matrix, and the corresponding Dynamic Functional Connectivity Vector (DFCV) samples were further extracted and collected. Finally, the DFCV samples were learned and classified by one sparse representation based method called Fisher Discriminative Dictionary Learning (FDDL). Total eight different whole-brain functional connectome patterns representing the dynamic brain states were obtained from this motor-related task experiment. The spatial distributions of functional connectivity strength showed obvious variance within different patterns. The pattern #1, pattern #2 and pattern #3 covered most of the samples (77.6%) and the similarities between each of them and the average static whole-brain functional connectivity strength matrix were obviously higher than other five patterns. Furthermore, the brain states were found to transfer from one pattern to another according to certain rules. The experimental results show that the proposed analysis method combining whole-brain DFC and FDDL learning is effective for describing and characterizing the dynamic brain states during task brain activity. It provides a foundation for exploring the dynamic information processing mechanism of the brain.

Since the face images might be not over-complete and they might be also corrupted under different viewpoints or different lighting conditions with noise, an efficient and effective method for Face Recognition (FR) was proposed, namely Robust Principal Component Analysis with Collaborative Representation based Classification (RPCA_CRC). Firstly, the face training dictionary D₀ was decomposed into two matrices as the low-rank matrix D and the sparse error matrix E; Secondly, the test image could be collaboratively represented based on the low-rank matrix D; Finally, the test image was classified by the reconstruction error. Compared with SRC (Sparse Representation based Classification), the speed of RPCA_CRC on average is 25-times faster. Meanwhile, the recognition rate of RPCA_CRC increases by 30% with less training images. The experimental results show the proposed method is fast, effective and accurate.

The existing passenger flow counting sensors with PolyVinyliDene Fluoride (PVDF) piezoelectric material exist the lack of accuracy caused by erroneous counting and missing counting, which has characteristics of low cost and resistance to wear and tear. In order to solve the problem, a passenger counting system based on PVDF gait intelligent detection technology was proposed. The ANSYS software was applied to carry out stress analysis of passengers' gait stepping on and off the bus and observe the distribution of the PVDF piezoelectric signal. The multi-input signal conditioning circuit was designed to acquire multi-channel plantar signal. Combined with signal processing algorithm, the sensor mechanical structure and people-counting system on buses were introduced by Laboratory Virtual Instrument Engineering Workbench (LabVIEW).The experimental results indicate that the proposed system improves the precision in comparison with the existing PVDF passenger flow counting sensors, reduces the cost in comparison with video image counting and human body infrared detection technology, and the average counting error is 5.3%.The proposed system has high practicality and can be widely used in Chinese public transport buses.

To improve the integrity, confidentiality and privacy of network-coding-based data transmission, a secure protection mechanism combined digital watermarking, stack shuffle and Message Authentication Code (MAC) was proposed. In this mechanism, the confidentiality and privacy were provided by mixing up messages using exclusive OR (XOR) encryption and stack shuffle. Furthermore, the confidentiality was enhanced by randomly inserting MACs into mixed messages with digital watermarking technique. And the integrity was provided by checking MACs on intermediate nodes during transmitting. The simulation results show that the spread hops of polluted information were effectively reduced by using this mechanism (less than 1.5). The collusion probability was less than 0.1 even if there were 25 collusion attackers and the size of key pool was 100. Both of theoretical analysis and simulation experiment demonstrate that the proposed mechanism can defend eavesdropping attacks, flow analysis attacks and polluting attacks with low expense.

To tackle the higher requirement of mobile network for movie service system and the lack of description of movie domain knowledge, the necessity and feasibility of establishing the Movie Ontology (MO) were illustrated. Firstly, the objects and components of MO were summarized, and the principle and method for building the MO model were also put forward, with using the Web Ontology Language (OWL) and Protege 4.1 to build the model. After that, the concrete representation of the class, property, individual, axioms and inference rules in the MO were explained. Finally, the consistency of MO was analyzed, including the consistency analysis of relationship between classes and the consistency analysis based on axioms.

K-Shortest-Paths (KSP) problem is the optimization issue in international flight route network. With the analysis on the international flight route network and KSP algorithm, the typical Yen algorithm solve KSP problem was investigated. To resolve the problem that Yen algorithm occupied much time in solving the candidate paths, an improved Yen algorithm was proposed. The improved Yen algorithm was set up by using the heuristic strategy of A* algorithm, which reduced the time to generate candidate paths, thereby, the search efficiency was improved and the search scale was reduced. The simulation results of international flight route network example show that the improved Yen algorithm can quickly solve KSP problem in international flight route network. Compared with the Yen algorithm, the efficiency of the proposed algorithm is increased by 75.19%, so it can provide decision support for international flight route optimization.

To improve the prediction ability of image memorability, a method for automatically predicting the memorability of an image was proposed by using visual saliency entropy and improved Object Bank feature. The proposed method improved the traditional Object Bank feature and extracted the visual saliency entropy feature. Then a prediction model of image memorability was constructed by using Support Vector Regression (SVR). The experimental results show that the correlation coefficiency of the proposed method is three percentage higher than the state-of-the-art method. The proposed model can be used in image memorability prediction, image retrieval ranking and advertisement assessment analysis.

Electronic Paper Display (EPD) can exhibit good comfortability in reading, but it has a critical drawback - slow refresh, which will be overcome by optimizing the design of the display's driver software. A tri-buffer-based architecture as well as its design methodology for the EPD driver software was proposed in this paper. Also an e-reader integrating the EPD driver in it was implemented to verify the design. Compared with the traditional dual-buffer architecture, the proposed tri-buffer scheme set an additional memory area to keep the EPD data frame. Test results show that the driver software works well in a real device without screen flicker and can help the display to achieve excellent performance.

To overcome the drawback of the existing immunity watermarking algorithm that the original image cannot be recovered exactly, a new reversible watermarking algorithm in wavelet domain based on immune watermarking framework was proposed. The algorithm has large embedding capacity and can recover the original image exactly. Meanwhile, the algorithm used histogram shifting method of reversible watermarking to embed recovery vector for recovering the original image precisely. By using the block size in the wavelet transform, the amount of the peak-zero point pair and the round amount in the histogram shifting, 〖JP2〗the control factor in the algorithm could be selected to decide the embedding depth. Then the algorithm could generate publishing image which has large distortion but also kept the main information of the original image. The scrambling and encryption methods were used in the algorithm so that the invaders could not recover the original image correctly by brute-force method without authorized file. The experimental results show that the algorithm can not only obtain the publishing image which has big differences with the original one, but also can recover the exact image when security permission is released.

The analysis results on PPStream-VOD System client behavior characteristics were presented in this paper. This study began from researching on peer-distributing protocol and the architecture of Buffer-Map based on passive measurement. A dedicated PPS-VOD crawler was deployed to capture clients’ Buffer-Map and study the characteristics of client watching behavior. By accurate data analysis, the client behavior was classified as Long-Smoother, Short-Smoother and Jumper. Then the proportion of three kinds of clients and their different watching behaviors were proposed. The concept watching viscosity was put forward to reveal the attraction of program to users, which is in direct proportion to average watching time, and in inverse proportion to slope of probabillty accumulation curve.

An image encryption algorithm based on inter-perturbations of dual chaotic systems was proposed for the possible degradation of low-dimensional chaotic system and the high computing work of high-dimensional chaotic system. Dual chaotic system was constructed through inter-perturbations of two simple Logistic mappings. And the most prominent feature of inter-perturbations between chaotic systems was that perturbation included constant perturbation and random perturbation simultaneously, which not only ensured the necessary system complexity, but also increased the range of system parameters. Taking dual chaotic inter-perturbed system as sequence key generator, an improved quantization method for converting chaotic sequence to binary sequence was put forward. Random testing and correlation analysis were done on binary sequence. The results show that the binary sequence has good pseudo-randomness and correlation, and appropriate to be encryption key. Simulation results of image encryption applied with the binary sequence also show that the binary sequence can cover up plaintext effectively and safely, and good encryption results are achieved.

Through the analysis of MPEG-4 video coding essential technology and principles,a video-medical system was developed,based on JMF and applied in the intellectualized community by using real-time,highly effective wireless video transmission features of MPEG-4.In video service system,JMF provides a unified construction and correspondence agreement to manage the gain,processing and the transmission of the time-based media.The adoption of real-time transmission agreement/Real-time transmission controlling agreement(RTP/RTCP) guarantees the video transmission process timeliness well,and introduces MPEG-4 compression algorithm plug-in unit in JMF to ensure the video compression.The compression rate of MPEG-4 compression algorithm is relatively higher and has strong network compatibility,which produces good effect in the procedure realization process.

Shot cut detection is one of the challenging problems in video auto-index and retrieval. A multi-scale algorithm of shot cut detection on MPEG compressed domain which analyzed the video stream on the scales of GOP, slot and B frames was proposed in this paper. The I frames in two adjacent GOP was examined to find if there were shot cut whitin the GOP; the area of the cut was located by analyzing the slot and the exact frame where the shot cut occurred was found by examining B frames between two reference frames.