To address problems of missing occluded objects, distant objects and objects in extreme weather scenarios when using lidar for object detection in autonomous driving, an attention-based object detection method with millimeter wave radar-lidar feature fusion was proposed. Firstly, the scan frame data of millimeter wave radar and lidar were aggregated into their respective labeled frames, and the points of millimeter wave radar and lidar were spatially aligned, then PointPillar was employed to encode both the millimeter wave radar and lidar data into pseudo images. Finally, the features of both millimeter wave radar and lidar sensors were extracted by the middle convolution layer, and the features maps of them were fused by attention mechanism, and the fused feature map was passed through a single-stage detector to obtain detection results. Experimental results on nuScenes dataset show that compared to the basic PointPillar network, the mean Average Precision (mAP) of the proposed attention fusion algorithm is higher, which performs better than concatenation fusion, multiply fusion and add fusion methods. The visualization results show that the proposed method is effective and can improve the robustness of the network for detecting occluded objects, distant objects and objects surrounded by rain and fog.

Since there are many troubles such as large quantity, uneven quality and poor credibility in getting helpful information and making effective decision for stakeholders, and the existing research work on credibility assessment mainly considers the sources of reviews and the support of reviews in form of votes, an assessment method on review credibility from perspective of intrinsic quality was proposed. That is, the credibility assessment of reviews was realized by integrating the ratings of reviewers, the support degree of reviews and the consistency in reviews, etc. Firstly, the pre-processing of review data was completed based on rule and method libraries. Then, the feature recognition and the feature value extraction and standardization were completed based on product feature library, generic dictionary, sentiment dictionary and method library. Finally, the credibility assessment of reviews was completed based on the established models. The experimental results verify the feasibility of this method, and it can be applied to assess the credibility of product reviews automatically on other e-commerce platforms.

To solve the problem of inconsistent data standards, uneven data quality, and compromised data security privacy in current data governance process, a new blockchain-based data governance collaboration method which integrates the characteristics of multi-party cooperation, security and reliability of blockchain was proposed and applied to the construction of data standards, the protection of data security, and the control of data sharing processes. Based on data governance requirements and blockchain characteristics, a collaborative data governance model was formed. A multi-collaborative data standard process, an efficient data standard construction and update mechanism, and secure data sharing and access control were developed afterwards. So this blockchain-based data governance collaboration method could be implemented to improve the efficiency and security of data standardization work. The experimental and analysis results show that the proposed method has a significant improvement in the efficiency of the standard term application time than the traditional data standard construction method. Especially in the big data environment, the application of the smart contract improves the time efficiency. The distributed storage of the blockchain provides powerful basis and guarantee for system security, user behavior traceability and auditing. Methods mentioned above provide a good application demonstration for data governance and a reference for the industry's metadata management, data standards sharing and application.

To solve privacy leakage and other security problems in Radio-Frequency Identification (RFID) tag authentication and tag ownership transfer, and to simplify the design of tag ownership transfer protocol, an RFID authentication protocol enabling tag ownership transfer was proposed for those tags that support Elliptic Curve Cryptography (ECC). The structure of the protocol is similar to the structure of the Diffie-Hellman logarithm, and tag privacy of the protocol is based on complexity to solve the computational Diffie-Hellman problem. Analysis of tag privacy and other security properties of the protocol were given, and comparisons between recent ECC-based authentication protocols and the proposed protocol were also given. The results show that the proposed protocol achieves best performance, under a comprehensive evaluation of supporting tag ownership transfer, tag computation cost, communication cost and tag privacy protection. In addition, a simplified version that realizes tag authentication to reader and is suitable for secure environments was also given.

Based on ARM920T kernel S3C2440, embedded Web services, QT technology and wireless networking technology, a smart home monitoring system was designed. The system was composed of a host of smart home, ZigBee/Wi-Fi wireless sensor control network and smart home client software. The hardware and software design of the host of smart home was completed:the embedded Linux operating system was transplanted in the ARM platform; the embedded Web services were established by using gSOAP tool; USB to serial driver and Wi-Fi wireless LAN (Local Area Network) driver were configured; ZigBee wireless sensor control network was formed, the program design of the coordinator node and terminal node was completed, the data communication protocol was made, and the client program was designed by using QT technology. Finally, the tests of establishment of ZigBee network, terminal nodes joining the network and sensor node data transmission were done. The test results show that the sensor nodes in the network can transmit the detection information to the coordinator, and the smart home client software can complete the remote monitoring and control of home environment through the host of smart home.

The Adaptive Generalized Total Variation (AGTV) model for image denoising has the shortages that it cannot locate image edge accurately and extract enough edge information. In order to improve the effectiveness and Peak Signal-to-Noise Ratio (PSNR) of image denoising, an Improved AGTV(IAGTV) model for image denoising was presented. On the one hand, another gradient calculating method with higher accuracy was adopted, in order to locate image edge more accurately than AGTV. On the other hand, for optimizing the filtering of image preprocess, the united Gauss-Laplace conversion which was good at image edge information detection was chosen to take place of Gaussian smoothing filter, so as to prevent edge information from reduction while denoising. Numerical simulation experiments show that the restored image PSNR of IAGTV was increased approximately by 1 dB than that of GTV with the fixed value p and at least 0.2 dB than that of AGTV. The experimental results show that IAGTV has good ability of image denoising.

For the problem of existing models' disability to analyze topic-sentiment evolution of microblogs, a Dynamic Topic Sentiment Combining Model (DTSCM) was proposed based on Topic Sentiment Combining Model (TSCM) and the emotional cycle theory. DTSCM could track the topic sentiment evolution trend and obtain the graph of topic sentiment evolution so as to analyze the evolution of topic and sentiment by capturing the topic and sentiment of microblogs in different time. The experimental results in real microblog corpus showed that, in contrast with state-of-the-art models Joint Sentiment/Topic (JST), Sentiment-Latent Dirichlet Allocation (S-LDA) and Dependency Phrases-Latent Dirichlet Allocation (DPLDA), the sentiment classification accuracy of DTSCM increased by 3.01%, 4.33% and 8.75% respectively,and DTSCM could obtain topic-sentiment evolution of microblogs. The proposed approach can not only achieve higher sentiment classification accuracy but also analyze topic-sentiment evolution of microblog, and it is helpful for public opinion analysis.

With reference to the properties of multiscale and shift invariance of nonsubsampled Contourlet transform, and concerning the characteristics of Computed Tomography (CT) and Magnetic Resonance Imaging (MRI) images, a medical images fusion method was proposed．The proposed method fused the low frequency subband and high frequency subband of these medical images separately by the regional feature strategy. The paper introduced the judgment criteria of images fusion and expatiated on the principle and implementation of Nonsubsampled Contourlet Transform (NSCT). And this gave the subjective judgment and numeric measurement of the fusion images based on visual effect and information indexes. To evaluate the performance of the proposed algorithm, the authors compared the results with those of the algorithms, such as wavelet transform and Contourlet transform. The CT and MRI images simulation results of mandibular system indicate that the proposed method outperforms the others in terms of both visual quality and objective evaluation criteria, while it can integrate and maintain much more effective and detailed information as well.

Concerning the faults of local extremum in image registration based on mutual information, a new medical image registration method based on Powell and improved genetic algorithm was proposed in this paper. It put forward an improved method regarding the shortcomings of the standard genetic algorithm, such as slow convergence and prematurity that will result in artifacts, and generated the iteration individual by Logistic chaos map. This method utilized the multi-resolution analysis strategy and searched for the optimal of the objective function by this hybrid optimized algorithm in the lowest resolution image level. Then it continued the optimization course and accomplished the image registration by this optimal data with the Powell algorithm. The experimental results indicate that this algorithm can effectively improve the image registration velocity and avoid local extremum of the operator while getting better performance of image precision in contrast to the Powell algorithm and unimproved genetic algorithm.

The author presented a snow simulation model in the 3D meteorological disaster analysis system. It could help analysts prevent snow storm and avalanches. First, the author classified the pixels of the satellite remote sensing images by their color vectors’ distance. Each of the colors represents a landscape. Then the author derived a formula by using mechanics. This formula was used to determine whether or not the snow could be covered on different regions. These regions had same landscape but different topography. If the regions were determined covered by snow according to the formula, the snow color was used to fill the regions. After this process, the satellite remote sensing images were used as textures in the 3D meteorological disasters analysis system. The experimental results show that snow effects can be produced by the snow simulation model.

Concerning low classification accuracy and high false alarm rate of current intrusion detection models, an intrusion detection classification model based on Landmark ISOmetric MAPping (LISOMAP) and Deep First Search (DFS) Relevant Vector Machine (RVM) was proposed. The LISOMAP was adopted to reduce the dimension of the training data, and RVM based on the DFS was used for classification detection. Compared with the Principal Components Analysis (PCA)-Supported Vector Machine (SVM), the experimental results indicate that the LISOMAP-DFSRVM model has lower false alarm rate with almost the same detection rate.

To increase face recognition speed and reduce RAM occupancy on a large-scale database, a high-efficient face recognition method was proposed. By introducing the technique of Discrete Cosine Transform (DC) image data compression into the face feature database compression, the database was compressed in multi-level compression ratio to generate few compressed sub-databases. Then, according to the order from high-to-low compression ratio, the rough face recognition was implemented on these sub-databases one by one. In the meantime, the recognition scope was narrowed down progressively according to previous recognition results. Finally, the accurate face recognition was carried out on the original uncompressed database in a very small range. Compared to traditional method, the experimental results show that the recognition time is reduced to 29.2%, RAM occupancy is reduced to 10.2%, and hard disk resource consumption is increased by only 11%, and the recognition rate is not significantly reduced.

Deploying sensor nodes’ has to meet certain coverage and connectivity in the energy constrained wireless sensor networks. Concerning this, a new square regionbased coverage and connectivity probability computational method was proposed in this paper. The new method can not only depict the relations among coverage rate, connectivity rate, the number of nodes, the sensing (communication) range of nodes and the size of network, but also calculate the number of sensor nodes scattered for maintaining certain coverage rate and connectivity rate. The simulation results show that the errorrate of deployment is less than the value obtained from the theoretical analysis.

Ray tracing algorithm needs a great deal of computation time. To improve its efficiency, a parallel ray tracing model, based on the image theory and the framework of .NET Remoting, was proposed in this paper. It scheduled and started the remote computing object, and assigned masks dynamically through computing the completion description with the sub-tree of secondary visual source. The comparison among numeric results indicates that the parallel model considerably reduces the computation time without losing precision. It can be applied to urban microcellular environments with arbitrary building layouts and heights.

Suspect fingerprint images acquired from criminal scenes are often of low quality, or incomplete. The existing algorithms/methods found in literature are not suitable for enhancing these fingerprint images, because these algorithms/methods only perform enhancement once. In this paper, a new method for fingerprint image enhancement was proposed. The proposed method enhanced fingerprint image in Fourier domain and spatial domain by using iterative and wavelet fusion technique. Experimental results indicate that the proposed method significantly improves the performance of fingerprint enhancement — not only was the ridge definition enhanced, but also the ridge fracture was connected and the incompleteness was mended. Therefore, the proposed method is suitable for enhancing low quality fingerprint images.

An octree based on the rapid k nearest search was presented for scattered points from 3D scanner. Through the establishment of bounding box on point sets, octree was used to record segmentation process, so that the search of point neighbors was limited to the bounding box of sample points and its neighbor, and through pruning strategies to further narrow the scope of the search.The experiments on a great deal of real data show that the search speed can be well improved.

According to more merits in some aspects over real wavelet transform such as shift invariability, more directions and precise information in phase space,complex wavelet transform can improve the ability to image denosing.In this paper combined with H-Curve criterion Dual-Tree complex wavelet transform was applied to image denosing. The H-Curve criterion does not need know the noise standard deviation firstly,and it can dispose many kinds of noise practically and produce a better effect in vision over smooth image brought up by many methods presently. The typical denosing test shows that the method produces a better effect over multitudinous methods presently in many aspects such as the ability to denosing, the effect in vision and universality in choosing a threshold.