In the real-time and complex network environment， how to motivate workers to participate in tasks and obtain high-quality perception data is the focus of spatio-temporal crowdsourcing research. Based on this， a spatio-temporal crowdsourcing’s online incentive mechanism based on quality perception was proposed. Firstly， in order to adapt to the real-time characteristics of spatio-temporal crowdsourcing， a Phased Online selection of workers Algorithm （POA） was proposed. In this algorithm， the entire crowdsourcing activity cycle was divided into multiple stages under budget constraints， and workers were selected online in each stage. Secondly， in order to improve the accuracy and efficiency of quality prediction， an Improved Expected Maximum （IEM） algorithm was proposed. In this algorithm， the task results submitted by workers with high reliability were given priority in the process of algorithm iteration. Finally， the effectiveness of the proposed incentive mechanism in improving platform utility was verified by comparison experiments on real datasets. Experimental results show that in terms of efficiency， compared with the Improved Two-stage Auction （ITA） algorithm， the Multi-attribute and ITA （M-ITA） algorithm， Lyapunov-based Vickrey-Clarke-Groves （L-VCG） and other auction algorithms， the efficiency of POA has increased by 11.11% on average， and the amount of additional rewards for workers has increased by 12.12% on average， which can encourage workers to move to remote and unpopular areas； In terms of quality estimation， the IEM algorithm has an average improvement of 5.06% in accuracy and 14.2% in efficiency compared to other quality estimation algorithms.

The low-rank constraint model based on traditional Nuclear Norm Minimization(NNM)tends to cause local texture detail loss in the denoising of Low-Dose CT(LDCT)image. To tackle this issue,a regional-content-aware weighted NNM algorithm was proposed for LDCT image denoising. Firstly,a Singular Value Decomposition(SVD)based method was proposed to estimate the local noise intensity in LDCT image. Then,the target image block matching was performed based on the local statistical characteristics. Finally,the weights of the nuclear norms were adaptively set based on both the local noise intensity of the image and the different singular value levels,and the weighted NNM based LDCT image denoising was realized. The simulation results illustrated that the proposed algorithm decreased the Root Mean Square Error(RMSE)index by 30. 11%,14. 38% and 8. 75% respectively compared with the traditional NNM,total variation minimization and transform learning algorithms,and improved the Structural SIMilarity(SSIM)index by 34. 24%,23. 06% and 11. 52% respectively compared with the above three algorithms. The experimental results on real clinical data illustrated that the mean value of the radiologists' scores of the results obtained by the proposed algorithm was 8. 94,which is only 0. 21 lower than that of the corresponding full dose CT images,and was significantly higher than those of the traditional NNM,total variation minimization and transform learning algorithms. The simulation and clinical experimental results indicate that the proposed algorithm can effectively reduce the artifact noise while preserving the texture detail information in LDCT images.

Aiming at the insufficiency of feature extraction in the existing Content-Based Medical Image Retrieval (CBMIR) algorithms, which resulted in imperfect semantic information representation and poor image retrieval performance, an algorithm based on multi-scale dense network was proposed. Firstly, the size of pulmonary nodule image was reduced from 512×512 to 64×64, and the dense block was added to solve the gap between the extracted low-level features and high-level semantic features. Secondly, as the information of pulmonary nodule images extracted by different layers in the network was varied, in order to improve the retrieval accuracy and efficiency, the multi-scale method was used to combine the global features of the image and the local features of the nodules, so as to generate the retrieval hash code. Finally, the experimental results show that compared with the Adaptive Bit Retrieval (ABR) algorithm, the average retrieval accuracy for pulmonary nodule images based on the proposed algorithm under 64-bit hash code length can reach 91.17%, which is increased by 3.5 percentage points; and the average time required to retrieve a lung slice is 48 μs. The retrieval results of the proposed algorithm are superior to other comparative network structures in expressing rich semantic features and retrieval efficiency of images. The proposed algorithm can contribute to doctor diagnosis and patient treament.

An improved Artificial Bee Colony (ABC) algorithm for function optimization based on central solution was proposed to solve the problem of poor local searching capacity and low accuracy of conventional ABC algorithm. The algorithm combined the advantage of the central solution, which was introduced into the local search process of onlooker bees. Onlooker bees chose some nectar sources with better fitness values using roulette, did the further local optimization based on central solution and updated the value of each dimension of nectar source in every iteration. In order to verify the validity of the proposed algorithm, six standard functions were selected to simulate and compare with the other tow algorithms including ABC algorithm and Best-so-far ABC algorithm, the proposed algorithm greatly improved the quality of solution and reached theoretical optimal value especially for Rastrigin function. The results show that the proposed algorithm has significant improvement on solution accuracy and convergence rate.

Focused on the traditional methods of feature selection for brain functional connectivity matrix derived from Resting-state functional Magnetic Resonance Imaging (R-fMRI) have feature redundancy, cannot determine the final feature dimension and other problems, a new feature selection algorithm was proposed. The algorithm combined Random Forest (RF) algorithm in statistical method, and applied it in the identification experiment of schizophrenic and normal patients, according to the features are obtained by the classification results of out of bag data. The experimental results show that compared to the traditional Principal Component Analysis (PCA), the proposed algorithm can effectively retain important features to improve recognition accuracy, which have good medical explanation.

Building an interpretable and large-scale protein-compound interactions model is an very important subject. A new chemical interpretable model to cover the protein-compound interactions was proposed. The core idea of the model is based on the hypothesis that a protein-compound interaction can be decomposed as protein fragments and compound fragments interactions, so composing the fragments interactions brings about a protein-compound interaction. Firstly, amino acid oligomer clusters and compound substructures were applied to describe protein and compound respectively. And then the protein fragments and the compound fragments were viewed as the two parts of a bipartite graph, fragments interactions as the edges. Based on the hypothesis, the protein-compound interaction is determined by the summation of protein fragments and compound fragments interactions. The experiment demonstrates that the model prediction accuracy achieves 97% and has the very good explanatory.

An improved multi-objective Cuckoo Search Algorithm (CSA) was proposed to overcome basic multi-objective CSA's default of low convergence speed in the later period and low solution quality when it was used to solve the multi-resource leveling problem. Firstly, a non-uniform mutation operator was embedded in the basic multi-objective cuckoo search to make a perfect balance between exploration and exploitation. Secondly, a differential evolution operator was employed for boosting cooperation and information exchange among the groups to enhance the convergence quality. The simulation test illustrates that the improved multi-objective CSA outperforms the basic multi-objective CSA and Vector Evaluated Particle Swarm Optimization Based on Pareto (VEPSO-BP) algorithm when global convergence is considered.

The hard disk of cloud computing platform is not reliable. This paper proposed to use Self-Monitoring Analysis and Reporting Technology (SMART) log to predict hard disk failure based on Classification using lOcal clusterinG with Over-Sampling (COG-OS) framework. First, faultless hard disks were divided into multiple disjoint sample subsets by using DBScan or K-means clustering algorithm. And then these subsets and another sample set of faulty hard disks were mixed, and Synthetic Minority Over-sampling TEchnique (SMOTE) was used to make the overall sample set tend to balance. At last, faulty hard disks was predicted by using LIBSVM classification algorithm. The experimental results show that the method is feasible. COG-OS improves SMOTE+Support Vector Machine (SVM) on faulty hard disks' recall and overall performance, when using K-means method to divide samples of faultless hard disks and using LIBSVM method with Radial Basis Function (RBF) kernel to predict faulty hard disks.

When the tracked targets get seriously obscured, temporarily leave the tracking screen or have significant displacement variation, adjoining interval updating algorithm based on on-line Boosting will lead to the error accumulation thus producing the drift or even tracking failure. Therefore, a reformative target tracking method based on on-line Boosting was proposed. The classifier feature library was updated by using on-line Boosting algorithm, and the threshold was dynamically renewed by using Kalman filter, hence the system could automatically capture the local features and apply corresponding adjustment to the value of threshold according to the tracking confidence of the object. When the confidence of the moving target was less than the lower threshold value, Blob tracking methodology would be applied. It processed as follows: the target was segmented into many regions according to the similarity of both color and space, and each single region contained the information of region number, location and size. One of the regions would be randomly selected into an on-line Boosting tracking module for testing, and the switch to the adjacent region by applying update algorithm for tracking would not happen unless the captured confidence level reached the upper threshold. Results of tests on different video sequences show that the proposed algorithm is capable of speedily and accurately capturing the target object real-time and holding a better robustness in comparison of the traditional on-line Boosting algorithm and other tracking algorithms.