The goal of Camouflage Object Segmentation （COS） is to detect hidden objects from the background. In recent years， Camouflage Object Detection （COD） based on Convolutional Neural Network （CNN） has developed rapidly， but there is still a problem that the complete object cannot be accurately detected in scenes with highly similar foreground/background. For the above problem， a COS method based on Channel Attention （CA） and edge fusion， called CANet （Network based on Channel Attention and edge fusion）， was proposed to obtain a complete segmentation result with clearer edge details of camouflage objects. Firstly， the SE （Squeeze-and-Excitation） attention was introduced to extract richer high-level semantic features. Secondly， an edge fusion module was proposed to restrain interference in low-level features and make full use of edge details information of the image. Finally， a channel attention module based on depthwise separable convolution was designed to gradually integrate cross-level multi-scale features in a top-down manner， which further improved detection accuracy and efficiency. Experimental results on multiple public COD datasets show that compared to eight mainstream methods such as SINet （Search Identification Net）， TINet （Texture-aware Interactive guidance Network） and C²FNet （Context-aware Cross-level Fusion Network）， CANet performs better and can obtain rich camouflage objects’ internal and edge detail information. Among them， CANet improves the structure-measure index by 2.6 percentage points compared to SINet on the challenging COD10K dataset. CANet has superior performance and is suitable for medical detection of lesion areas similar to human tissue， military detection of hidden targets， and other related fields.

In recent years， anchor-based 3D hand pose estimation methods are becoming popular， and Anchor-to-Joint （A2J） is one of the more representative methods. In A2J， anchor points are densely set on depth map， and neural network is used to predict offsets between anchor points and key points together with weights of anchor points； predicted offsets and weights are used to calculate the coordinates of key points in a weighted summation mode to reduce noise in network regression results. A2J methods are simple and effective， but they are sensitive to ill-suited network structure and prone to inaccurate regression due to loss function. Therefore， an improved network HigherA2J was proposed. Firstly， a single branch jointly predicted X， Y and Z offsets between anchors and key points to better utilize 3D characteristics of depth map； secondly， network branch structure was simplified to reduce network parameters； finally， the loss function for key point estimation was designed， combined with offset estimation loss， which improved the overall estimation accuracy effectively. Experimental results show the reductions in average hand pose estimation error of 0.32 mm， 0.35 mm and 0.10 mm compared to conventional A2J on three datasets NYU， ICVL and HANDS 2017 respectively.

Since the traditional wavelet threshold functions have some drawbacks such as the non-continuity on the points of threshold, and large deviation of estimated wavelet coefficient, distortion and Gibbs phenomenon occur after denoising. To overcome these drawbacks, an improved threshold function was proposed. Compared with the hard, soft threshold functions and the existing improved threshold function, the proposed function not only is easy to be calculated, but also has the superior mathematical characteristics.To verify its advantages, a series of simulation experiments were performed, the Peak Signal-to-Noise Ratio (PSNR) and Mean Squared Error (MSE) values were compared with other different denoising methods.The experimental results indicate that it is better than above mentioned denoising methods in both the visual effects and the performance of PSNR and MSE.

To solve the problem of large amount of calculation and nonlinear programming in the process of service composition optimization, a Cost Benefit Coefficient (CBC) approach was proposed for Web services composition reliability optimization in the situation of a given cost investment. First, the structure patterns of service composition and related reliability function were analyzed. Furthermore, the Web service composition method of reliability calculation was proposed and a nonlinear optimization model was established accordingly. And then the cost benefit coefficient was computed through the relationship between the cost and the reliability of component services, and the optimization schemes of Web service composition were decided. According to the nonlinear optimization model, the results of optimization were computed. Finally, given cost investment, the higher reliability of the approach to optimize the reliability of Web service composition was verified through the comparison of this approach and the traditional method on the reliable data of component service. The experimental results show that the proposed algorithm is effective and reasonable for reliability optimization of Web services composition.

The seizure detection is important for the localization and classification of epileptic seizures. In order to solve the problem brought by large amount of data and high feature space in EEG (Electroencephalograph) for quickly and accurately detecting the seizures, a method based on max-Relevance and Min-Redundancy (mRMR) criteria and Extreme Learning Machine (ELM) was proposed. The time-frequency measures by Short-Time Fourier Transform (STFT) were extracted as features, and the large set of features were selected based on max-relevance and min-redundancy criteria. The states were classified using the extreme learning machine, Support Vector Machine (SVM) and Back Propagation (BP) algorithm. The result shows that the performance of ELM is better than SVM and BP algorithms in terms of computation time and classification accuracy. The classification accuracy rate of interictal durations and seizures can reach more than 98%, and the computation efficiency is only 0.8s. This approach can detect epileptic seizures accurately in real-time.

Opportunistic network routing algorithm based on geographic location information in DIrection based Geographic routing scheme (DIG) has the problems of large delay and low success rate, which is due to that DIG algorithm makes the waiting time of the data in the cache too long and cannot guarantee the data-carrying node move to the destination node. To solve these problems, Geographic Routing algorithm based on Directional Data Transmission (GRDDT) was proposed. The algorithm used a new data forwarding mechanism and a more effective use of the neighbor list information, effectively avoiding the appearance of the above circumstances, so as to reduce data packet transmission delay and to improve the success rate. OPNET simulation results show that, the performance of transmission delay and success rate of GRDDDT algorithm are improved compared with DIG.

According to the theory of sparse representation, images can be sparse-represented by using an appropriately redundant dictionary. The completeness can enable using very few big coefficients to capture the important information of images, and cause more robust to noise. Regarding image de-noising, considering the human visual characteristics, this paper studied the effective representation of characteristics and edge information of noisy image based on complete dictionary. For more effective feature retaining of images, a method of feature-retaining de-noising via sparse representation was proposed, which made the Structural SIMilarity (SSIM) as fidelity measure of the information. The experimental results indicate that the proposed algorithm has a better efficiency of de-noising, enhances the capacity of retaining feature, and gets a better visual effect of de-noised image.

When single Field-Programmable Gate Array (FPGA) deals with the huge amounts of data in high-speed network, low efficiency problem occurs. According to dual priority schedule algorithm for multi-processor and high-speed data acquisition and processing model based on multi-FPGA, a dual priority dynamic scheduling algorithm was proposed based on multi-FPGA. For strong real-time periodic tasks set in low priority queue, the Earliest Deadline Critical Laxity (EDCL) scheduling algorithm was given to determine the priority of task according to the degree of relaxation of the tasks. If the task was not finished when the promotion time was up, it would be promoted to high priority queue. For soft real-time periodic tasks, an algorithm was put forward to assign the tasks to middle priority queue and schedule them by delaying the deadline of tasks to dynamic blur threshold. The experimental results show that the proposed algorithms can well schedule strong real-time periodic tasks and guarantee the priority execution of important tasks, and it can also reduce miss rate of soft real-time periodic tasks caused by preemption.

With the wide application of categorical-attribute dataset, the demand of obtaining the cluster boundary of categorical-attribute dataset becomes more and more urgent. In order to get cluster boundaries, the individual proposed a categorical-attribute data boundary detection algorithm: CBORDER(Categorical dataset BORDER detection algorithm) In this algorithm, firstly, initializing the center of cluster by using random allocation and utilizing boundary-degree to partition clusters, at the same time, we get the evidence of capturing boundary recorders. Then, basing on the evidence accumulation, we execute the above procedure repeatedly to acquire the boundaries of clusters at the end. Multi-experimental results demonstrate that CBORDER detect boundaries of the high dimension categorical data efficiently.

Concerning the problems of the blind source separation algorithm that is not suitable for the broadband, complex computation, and being acute to various Signal Noise Ratio (SNR), a beamspace Blind Source Separation (BSS) method based on sub-band decomposition was proposed. The method of sub-band decomposition was used to realize the expansion of blind source separation under broadband condition, and beam transforming was used to reduce the amount of calculation. Simulation results demonstrate that the proposed method exhibits good source signal estimation capability. Even without any prior knowledge, it has the same performance as Root-Music high resolution algorithm.