To address the issue of image authenticity verification in deepfake detection and model copyright protection， a high-quality and highly robust watermarking method for diffusion model output， DeWM （Decoder-driven WaterMarking for diffusion model）， was proposed. Firstly， a decoder-driven watermark embedding network was proposed to realize direct sharing of encoder and decoder features， so as to produce watermarks with high robustness and invisibility. Then， a fine-tuning strategy was designed to fine-tune the pre-trained diffusion model's decoder， and embed a specific watermark into all generated images， thereby achieving simple and effective watermark embedding without changing the model architecture and diffusion process. Experimental results show that compared with Stable Signature method on the MS-COCO dataset， when the watermark bit-length is increased to 64 bits， the proposed method has the Peak Signal-to-Noise Ratio （PSNR） and Structure SIMilarity （SSIM） of the generated watermarked images improved by 14.87% and 9.41%， respectively. Moreover， the average bit accuracy of watermark extraction under cropping， brightness adjustment and image reconstruction is enhanced by than 3%， which demonstrates significantly improved robustness.

Focused on the challenges of edge information loss and incomplete segmentation of large lesions in endoscopic semantic segmentation networks， a Boundary-Cross Supervised semantic Segmentation Network （BCS-SegNet） with Decoupled Residual Self-Attention （DRA） was proposed. Firstly， DRA was introduced to enhance the network’s ability to learn distantly related lesions. Secondly， a Cross Level Fusion （CLF） module was constructed to combine multi-level feature maps within the encoding structure in a pairwise way， so as to realize the fusion of image details and semantic information at low computational cost. Finally， multi-directional and multi-scale 2D Gabor transform was utilized to extract edge information， and spatial attention was used to weight edge features in the feature maps， so as to supervise decoding process of the segmentation network， thereby providing more accurate intra-class segmentation consistency at pixel level. Experimental results demonstrate that on ISIC2018 dermoscopy and Kvasir-SEG/CVC-ClinicDB colonoscopy datasets， BCS-SegNet achieves the mIoU （mean Intersection over Union） and Dice coefficient of 84.27%， 90.68% and 79.24%， 87.91%， respectively； on the self-built esophageal endoscopy dataset， BCS-SegNet achieves the mIoU of 82.73% and Dice coefficient of 90.84%， while the above mIoU is increased by 3.30% over that of U-net and 4.97% over that of UCTransNet. It can be seen that the proposed network can realize visual effects such as more complete segmentation regions and clearer edge details.

Aiming at the problem that the existing human skeleton-based action recognition algorithms cannot fully explore the temporal and spatial characteristics of motion, a human skeleton-based action recognition algorithm based on Spatiotemporal Attention Graph Convolutional Network (STA-GCN) model was proposed, which consisted of spatial attention mechanism and temporal attention mechanism. The spatial attention mechanism used the instantaneous motion information of the optical flow features to locate the spatial regions with significant motion on the one hand, and introduced the global average pooling and auxiliary classification loss during the training process to enable the model to focus on the non-motion regions with discriminability ability on the other hand. While the temporal attention mechanism automatically extracted the discriminative time-domain segments from the long-term complex video. Both of spatial and temporal attention mechanisms were integrated into a unified Graph Convolution Network (GCN) framework to enable the end-to-end training. Experimental results on Kinetics and NTU RGB+D datasets show that the proposed algorithm based on STA-GCN has strong robustness and stability, and compared with the benchmark algorithm based on Spatial Temporal Graph Convolutional Network (ST-GCN) model, the Top-1 and Top-5 on Kinetics are improved by 5.0 and 4.5 percentage points, respectively, and the Top-1 on CS and CV of NTU RGB+D dataset are also improved by 6.2 and 6.7 percentage points, respectively; it also outperforms the current State-Of-the-Art (SOA) methods in action recognition, such as Res-TCN (Residue Temporal Convolutional Network), STA-LSTM, and AS-GCN (Actional-Structural Graph Convolutional Network). The results indicate that the proposed algorithm can better meet the practical application requirements of human action recognition.

Lane line detection is an important part of intelligent driving system. The traditional lane line detection method relies heavily on manual selection of features, which requires a large amount of work and has low accuracy when it is interfered by complex scenes such as object occlusion, illumination change and road abrasion. Therefore, designing a robust detection algorithm faces a lot of challenges. In order to overcome these shortcomings, a lane line detection model based on deep learning instance segmentation method was proposed. This model is based on the improved Mask R-CNN model. Firstly, the instance segmentation model was used to segment the lane line image, so as to improve the detection ability of lane line feature information. Then, the cluster model was used to extract the discrete feature information points of lane lines. Finally, an adaptive fitting method was proposed, and two fitting methods, linear and polynomial, were used to fit the feature points in different fields of view, and the optimal lane line parameter equation was generated. The experimental results show that the method improves the detection speed, has better detection accuracy in different scenes, and can achieve robust extraction of lane line information in various complex practical conditions.

In view of the problem that it is difficult to accurately recognize the type of vehicle due to scale change and deformation under multiple angles, a fine-grained vehicle recognition model based on Multi-Scale Bilinear Convolutional Neural Network (MS-B-CNN) was proposed. Firstly, B-CNN was improved and then MS-B-CNN was proposed to realize the multi-scale fusion of the features of different convolutional layers to improve feature expression ability. In addition, a joint learning strategy was adopted based on center loss and Softmax loss. On the basis of Softmax loss, a category center was maintained for each category of the training set in the feature space. When new samples were added in the training process, the classification center distances of samples were constrained to improve the ability of vehicle recognition in multi-angle situations. Experimental results show that the proposed vehicle recognition model achieved 93.63% accuracy on CompCars dataset, verifying the accuracy and robustness of the model under multiple angles.

The self-driving mine locomotive needs to detect and locate pedestrians in front of it in the underground roadway in real-time. Non-visual methods such as laser radar are costly, while traditional visual methods based on feature extraction cannot solve the problem of poor illumination and uneven light in the laneway. To solve the problem, a pedestrian visual positioning algorithm for underground roadway based on deep learning was proposed. Firstly, the overall structure of the system based on deep learning network was given. Secondly, a multi-layer Convolutional Neural Network (CNN) for object detection was built to calculate the two-dimensional coordinates and the size of bounding box of pedestrians in visual field of the self-driving locomotive. Thirdly, the third-dimensional distance between the pedestrian in the image and the locomotive was calculated by polynomial fitting. Finally, the model was trained, verified and tested through real sample sets. Experimental results show that the accuracy of the proposed algorithm reaches 94%, the speed achieves 25 frames per second, and the distance detection error is less than 4%, thus efficient and real-time laneway pedestrian visual positioning is realized.

In view of the problem that the test cost is increased by a large number of test suites in regression testing, a test suite reduction method based on weak mutation criterion was proposed. Firstly, the relation matrix between test suites and mutation branches was obtained based on weak mutation criterion. Then, four invalid test requirements and subset test suites were reduced repeatedly. Finally, the current optimal test suite was selected by using artificial fish swarm algorithm, and the simplification and test suite selection operations were performed alternately until all the test requirements were covered. Compared with Greedy algorithm and HGS (Harrold-Gupta-Soff) algorithm on six classical programs, when using weak mutation criterion with no changing or slightly changing mutation score, the reduction rate was improved by 73.4% and 8.2% respectively, and the time consumption was decreased by 25.3% and 56.1% respectively. The experimental results show that the proposed method can effectively reduce the test suites and save the test cost in regression testing.

During underwater signal data transmission process, using Field Programmable Gate Array (FPGA) to calculate Cyclic Redundancy Check (CRC) code with traditional serial calculating method cannot meet the demand of fast computation; however, parallel checking method, which is much faster, has difficulty in practical engineering application because of programming complexity. In order to meet the demand of transmission speed, to eliminate programming difficulty and time waste, a method was proposed to automatically generate parallel CRC code for any length data frames by Matlab. It finished all the mathematical deductions based on matrix method and calculations with the help of Matlab and then generated parallel CRC calculating program which conforms to the Verilog HDL grammar rules. Finally, the CRC calculation program statements generated by Matlab were first simulated in Quartus II 9.0 and then demonstrated by data transmission experiments on a civil towed sonar system. The results prove the validity of the proposed method, its programming and generation can be finished in tens of seconds, and the CRC module can accurately figure out CRC code of every long data frame defined by transmission protocol within requested time.

To solve the user similarity between trajectories formed by mobility data, an algorithm based on Location Sequence Generalized Suffix Tree (LSGST) was proposed. First, the location sequence was extracted from mobility data. At the same time the location sequence was mapped to a string. The transformation from the processing of location sequence to the processing of string was completed. Then the location sequence generalized suffix tree between different users was constructed. The similarity was calculated in detail from the number of similar positions, longest common subsequence and the frequent common position sequence. The theoretical analysis and simulation results show that the proposed algorithm has ideal effect in terms of similarity measure. Besides, compared to the ordinary construction method, the proposed algorithm has low time complexity. In the comparison with dynamic programming and naive string-matching, the proposed algorithm has higher efficiency when searching for the longest common sub-string and frequent public position sequence. The experimental results indicate that the LSGST can measure the similarity effectively, meanwhile reduces the trajectory data when searching for the measurement index, and has better performance in time complexity.

The traditional testing process does not specifically consider the system performance. With the wide application of parallel testing method, more attention was paid to the system performance and data throughput capacity. Optimizing the software design with multithreading technology becomes an effective way to improve the performance of automatic test system. By modeling testing pipeline process, using asynchronous pipeline design patterns and combining task-oriented concepts, an available test system programming model was proposed. The experiment results prove that the model can significantly shorten the average test time in the ideal case of random input of test tasks. Applying this model to an instance of measuring characteristic parameters of Alternating Current (AC) contactor, the results further indicate that this model can significantly increase the flexibility of test configuration and avoid the complexity of multi-threaded programming.

Consultative Committee for Space Data System (CCSDS) protocol is the mainstream of international space-ground link standard for space communication. The reversing of unknown CCSDS protocol can be used in at least two areas: one is to analyze the unknown communication traffics; the other is to detect and analyze the network attack aiming at space station as well as other space entities which are networked for international space co-operation. Thus, a computer aided analytical system was designed to reverse unknown protocol based on CCSDS protocol standard framework, and the system included the architecture design and the workflow design. Moreover, to solve the problem of telegram clustering efficiency of iterative phylogenetic tree of unknown protocol in the workflow, an improved algorithm, called Feedback Dynamic Relaxation Factor-Affinity Propagation (FDRF-AP), was given to solve the unknown communication protocol reversing problem. The simulation results indicate that the algorithm enhances the efficiency of protocol reversing engineering.

Concerning the key problems of selecting threshold function in wavelet threshold denoising, in order to address the discontinuity of conventional threshold function and large deviation existing in the estimated wavelet coefficients, a continuous adaptive threshold function in the whole wavelet domain was proposed. It fully considered the characteristics of different sub-band coefficients in different scales, and set the energy of sub-band coefficients in different scales as threshold function's initial weights. Optimal weights were iteratively solved by using interval advanced-retreat method and golden section method, so as to adaptively improve approximation level between estimated and decomposed wavelet coefficients. The experimental results show that the proposed method can both efficiently reduce noise and simultaneously preserve the edges and details of image, also achieve higher Peak Signal-to-Noise Ratio (PSNR) under different noise standard deviations.

To solve the low efficiency and precision of the traditional methods, a near-duplicate image detection algorithm based on Bag-of-words and Hash coding was proposed in this paper. Firstly, a 500-dimensional feature vector was used to represent an image by Bag-of-words; secondly, feature dimension was reduced by Principal Component Analysis (PCA) and Scale-Invariant Feature Transform (SIFT) and features were encoded by Hash coding; finally, dynamic distance metric was used to detect near-duplicate images. The experimental results show that the algorithm based on Bag-of-words and Hash coding is feasible in detecting near-duplicate images. This algorithm can achieve a good balance between precision and recall rate: the precision rate can reach 90%-95%, and entire recall rate can reach 70%-80%.

To manage the risk of service, it is necessary to obtain stochastic character of Quality of Service (QoS) that is represented as accurate probability distribution. This paper presented an approach to estimate probability distribution of Web service QoS in the case of small number of samples. Using max entropy principle, the analytical formula of the probability density function can be obtained by transforming the probability distribution estimation problem into an optimal problem with constraints obtained from sampling QoS data. Then an algorithm to estimate parameters of the probability density function was designed. The experimental and simulation results based on real Web service QoS data show the effectiveness of the proposed approach for probability distribution estimation of different QoS attribute. The efficiency and feasibility of the distribution estimation algorithm have got validated by experiments too.

At present, most of the privacy protection schemes are based on sophisticate zero-knowledge protocol or bilinear mapping computation, so their efficiency is low. In order to address this problem, an anonymous negotiation credentials scheme was proposed based on ring signature. On the foundation of anonymous credentials framework, an efficient discrete logarithm based ring signature scheme was constructed to protect the negotiation credentials, compared with two schemes proposed by ZHANG, et al. (ZHANG MING-WU, YANG BO, ZHU SHENG-LIN, et al. Policy-Based Signature Scheme for Credential Privacy Protecting in Trust Negotiation. Journal of Electronics & Information Technology, 2009(1): 224-227) and LIU, et al. (LIU BAILING, LU HONGWEI, ZHAO YIZHU. An efficient automated trust negotiation framework supporting adaptive policies. Proceedings of the Second International Workshop on Education Technology and Computer Science. Washington, DC: IEEE Computer Society, 2010: 96-99) the proposed scheme is more efficient.