Aiming at the problems that the existing Deep Convolutional Neural Network (DCNN) have insufficient defect image feature extraction ability, few recognition types and long reasoning time in hydraulic tunnel defect recognition tasks, an autonomous defect recognition method based on dynamic feature distillation was proposed. Firstly, the deep curve estimation network was used to optimize the image to improve the image quality in low illumination environment. Secondly, the dynamic convolution module with attention mechanism was constructed to replace the traditional static convolution, and the obtained dynamic features were used to train the teacher network to obtain better model feature extraction ability. Finally, a dynamic feature distillation loss was constructed by fusing the discriminator structure in the knowledge distillation framework, and the dynamic feature knowledge was transferred from the teacher network to the student network through the discriminator, so as to achieve the high-precision recognition of six types of defects while significantly reducing the model reasoning time. In the experiments, the proposed method was compared with the original residual network on a hydraulic tunnel defect dataset of a hydropower station in Sichuan Province. The results show that this method has the recognition accuracy reached 96.15%, and the model parameter amount and reasoning time reduced to 1/2 and 1/6 of the original ones respectively. It can be seen from the experimental results that fusing the dynamic feature distillation information of the defect image into the recognition network can improve the efficiency of hydraulic tunnel defect recognition.

To solve the problems of multiple network parameters and high computational complexity in image super-resolution reconstruction of deep network architecture, an image super-resolution reconstruction method based on accelerated residual network was proposed. Firstly, a residual network was constructed to reconstruct the high-frequency residual information between low-resolution image and high-resolution image, so as to reduce the deep network transmission process of redundant information and improve the reconstruction efficiency. Secondly, the dimensionality of the extracted low-resolution feature map was reduced by the feature shrinking layer to realize fast mapping with fewer network parameters. Thirdly, the dimensionality of the high-resolution feature map was increased by the feature expanding layer to reconstruct the high-frequency residual information with the rich information. Finally, the residual and low-resolution images were summed to obtain the reconstructed high-resolution image. Experimental results show that the Peak Signal-to-Noise Ratio (PSNR) and the Structural SIMilarity (SSIM) mean results obtained by the proposed method are 0.57 dB and 0.013 3 higher than those obtained by Super-Resolution using Convolutional Neural Network (SRCNN) respectively, and 0.45 dB and 0.006 7 higher than those obtained by Intermediate Supervision Convolutional Neural Network (ISCNN). In terms of reconstruction speed, using dataset Urban100 as example, the proposed method is 1.5 to 42 times faster than the existing methods. In addition, when this method is applied to the super-resolution reconstruction of motion blur images, it has the performance better than image Super-Resolution using Very Deep convolutional network (VDSR). The proposed method achieves better reconstruction quality with fewer network parameters and provides a new idea for image super-resolution reconstruction.

Thoracic Aortic Aneurysm and Dissection (TAAD) is one of the life-threatening cardiovascular diseases, and the histological changes of Medial Degeneration (MD) have important clinical significance for the diagnosis and early intervention of TAAD. Focusing on the issue that the diagnosis of MD is time-consuming and prone to poor consistency because of the great complexity in histological images, a deep learning based classification method of histological images was proposed, and it was applied to four types of MD pathological changes to verify its performance. In the method, an improved Convolutional Neural Network (CNN) model was employed based on the GoogLeNet. Firstly, transfer learning was adopted for applying the prior knowledge to the expression of TAAD histopathological images. Then, Focal loss and L2 regularization were utilized to solve the data imbalance problem, so as to optimize the model performance. Experimental results show that the proposed model is able to achieve the average accuracy of four-class classification of 98.78%, showing a good generalizability. It can be seen that the proposed method can effectively improve the diagnostic efficiency of pathologists.

At present, the deployment of nodes in wireless sensor network mainly adopts the algorithm based on Voronoi diagram. In the process of deployment using Voronoi algorithm, due to the large number of nodes involved in the deployment and the high complexity of the algorithm, the iteration time of the algorithm is long. In order to solve the problem of long iteration time in node deployment, a Deployment Algorithm based on Basic Architecture (DABA) was proposed. Firstly the nodes were combined into basic architectures, then center position coordinates of the basic architecture were calculated, finally the node deployment was performed by using Voronoi diagram. The algorithm was still able to realize the deployment effectively under the condition that there were obstacles in the deployment area. The experimental results show that DABA can reduce the deployment time by two thirds compared with the Voronoi algorithm. The proposed algorithm can significantly reduce the iteration time and the complexity of the algorithm.

Photo observation and identification is often influenced by insufficient light and unsuitable shooting angle when taking photos. In order to solve this problem, an image restoration method based on nonlinear brightness-lifting model was proposed. Although the existing nonlinear brightness enhancement method can improve the brightness of the backlight area, distortion still occurs in the highlighted area due to excessive promotion. On the basis of the existing image processing algorithm, a new adaptive backlight images restoration method based on nonlinear brightness improvement model was proposed. Image segmentation processing and logarithmic function were used to enhance image brightness, in which the threshold was determined by Otsu threshold processing, and the adjustment coefficient in the transition function was calculated by the ratio between pixels of backlight area and total pixels. Simulation results show that, compared with the method of using the logarithmic function conversion relation and adjusting the image brightness in the HSI space model, the proposed method not only improves the image quality and preserves the nature of image without distortion, but also has a good improvement in performance.

To help users to quickly retrieve the interesting game strategies, a knowledge driven automatic annotating algorithm for game strategies was proposed. In the proposed algorithm, the game domain knowledge base was built automatically by fusing multiple sites that provide information for each game. By using the game domain vocabulary discovering algorithm and decision tree classification model, game terms of the game strategies were extracted. Since most terms existing in the strategies in the form of abbreviation, the game terms were finally linked to knowledge base to generate the full name semantic tags for them. The experimental results on many games show that the precision of the proposed game strategy annotating method is as high as 90%. Moreover, the game domain vocabulary discovering algorithm has a better result compared with the n-gram language model.

The distributed state of noise existing among different bands of hyperspectral remote sensing image is complex, so the traditional denoising methods are hard to achieve the desired effect. In order to solve this problem, based on Principal Component Analysis (PCA), a novel denoising method for hyperspectral data was proposed combining with noise estimation and dictionary learning. Firstly, a group of the principal component images were achieved from the original hyperspectral data by using the PCA transform, which were divided into clear image group and noisy image group according to the corresponding energy. Then, according to any band image from noisy hyperspectral data, the noise standard deviation of the image was estimated via a noise estimation method based on Singular Value Decomposition (SVD). Meanwhile, combining this noise estimation method with denoising method via K-SVD dictionary learning, a new dictionary learning denoising method with adaptive noise estimation characteristics was proposed and applied to denoise those images from noisy image group with low energy where noise mainly existed. Finally, the final denoising image was obtained by weighted fusion according to the corresponding energy of each principal component image. The experimental results on simulated and real hyperspectral remote sensing data show that, compared with PCA, PCA-Bish and PCA-Contourlet, the Peak Signal-to-Noise Ratio (PSNR) of the image denoised by the proposed algorithm is improved by 1-3 dB, and more detailed information and better visual effect of the denoised image by the proposed method are achieved.

Aiming at real-time requirements and elimination ghost produced by moving object in video stitching, a method based on circular Region Of Interest (ROI) image registration was proposed by using the simplified process and Graphics Processing Unit (GPU) acceleration. Firstly, the feature extraction only occured in the ROI area, which improved the detection speed and the feature matching accuracy. Secondly, to further reduce the time cost and meet the real-time requirements for video processing, two strategies were used. On one hand, only the first frame was used for matching, while the subsequent frames used the same homography matrix to blend. On the other hand, GPU was adopted to realize hardware acceleration. Besides, when there are dynamic objects in the field of view, the graph-cut and multi-band blending algorithms were used for image blending, which can effectively eliminate ghost image. When stitching two videos of 640×480, the processing speed of the proposed method was up to 27.8 frames per second. Compared with the Speeded Up Robust Features (SURF) and Oriented features from Accelerated Segment Test (FAST) and Rotated BRIEF (ORB), the efficiency of the proposed method was increased by 26.27 times and 11.57 times respectively. Experimental results show the proposed method can be used to stitch multi-channel videos into a high quality video.

In view of significance of color image recognition, the method of color image recognition based on data of image features and Deep Belief Network (DBN) was presented. Firstly, data field of color image was constructed in accord with human visual characteristics; secondly, wavelet transforms was applied to describe multi-scale feature of image; finally, image recognition could be made by training unsupervised DBN.The experimental results show that compared with the methods of Adaboost and Support Vector Machine(SVM),classification accuracy is improved by 3.7% and 2.8% respectively and better image recognition is achieved by the proposed method.

Traditional trust and reputation models do not pay enough attention to nodes'deceit in recommendation, so their reputation evaluation may be affected by malicious nodes' collusion. A trust and reputation model named Enhanced Attack Resistible Ant-based Trust and Reputation Model (EAraTRM) was proposed, which is based on ant colony algorithm. Node recommendation behaviors analysis and adaptive mechanism to malicious nodes density were added into reputation evaluation of EAraTRM to overcome the shortage of traditional models. Simulation experiments show that EAraTRM can restrain the collusion of malicious nodes, and give more accurate reputation evaluation results, even when 90% nodes in a network are malicious and the comparison models have failed.

To solve the sensor node localization problem of Wireless Sensor and Actor Network (WSAN), a range-based localization algorithm with virtual force in WSAN was proposed in this paper, in which mobile actor nodes were used instead of Wireless Sensor Network (WSN) anchors for localization algorithm, and Time Of Arrival (TOA) was combined with virtual force. In this algorithm, the actor nodes were driven under the action of virtual force and made themself move close to the sensor node which sent location request, and node localization was completed by the calculation of the distance between nodes according to the signal transmission time. The simulation results show that the localization success rate of the proposed algorithm can be improved by 20% and the average localization time and cost are less than the traditional TOA algorithm. It can apply to real-time field with small number of actor nodes.

CT image denoising restoration is a basic procedure in medical image processing. Cartoon-texture decomposition method was extended in order to resolve the problems of computational difficulty and low precision while applying cartoon-texture models in medical image denoising. First, the structure of cartoon-texture model was described by curvelet transform. Second, the texture of cartoon-texture decomposition was described using more sparse dual-tree complex wavelet transform. Third, an image cartoon images-texture model was established by combining curvelet transform and sparse representation. The algorithms of cartoon-texture model were discussed at last. The simulation experimental results show that the new method can effectively resolve the problem of large amount of iterative calculation using medical image denoising algorithm, and the image quality after processing can be improved as well.

The Wireless Transport Layer Security (WTLS) handshake protocol was built based on digital certificate mechanism. However, there exist several flaws in WTLS. For example, both the communication and computation overload are high. Moreover, it does not verify the server certificate on-line. In order to solve these issues, an improved WTLS handshake protocol based on Identity-based Cryptosystem (IBC) was proposed. It is constructed based on ID, and IDs are exchanged between server and client instead of certificates. Identity-based Encryption (IBE), Identity-based Signature (IBS) and Identity-based Authenticated Key Agreement (IBAKA) were adopted to implement security functions of encryption, signature and key agreement respectively. Sender's ID information was embedded into encryption key computation, which can be used to authenticate the source of message. The analysis on security and efficiency shows that the efficiency of wireless communication is improved without security loss.

In order to meet the demands for both availability and processing speed, in reference to dark channel prior estimation, an image restoration algorithm based on Approximate Point Spread Function Examining (APEX) algorithm commonly used in image deblurring was proposed. Meanwhile, because different-sized images under different weather conditions have different APEX parameters, the APEX parameter value was adjusted dynamically according to sandstorm and fog degree. Furthermore, unlike other multiple images methods, the proposed algorithm needs only one image to be the input. The experimental results show that the proposed algorithm is effective in restoring images. By using color constancy algorithm, the source color components were balanced; furthermore, the visual effect of images was enhanced.