The Aspect-Based Sentiment Analysis （ABSA） task aims to identify the sentiment polarity of a specific aspect. However， the existing related models lack the short-distance constraints on the context of the aspect word for the natural sentences with uncertain structure， and easily ignore the syntactic relations， so it is difficult to accurately determine the sentiment polarity of the aspect. Aiming at the above problems， an ABSA model with Embedding Different Neighborhood Representations （EDNR） was proposed. In this model， on the basis of obtaining the word order information of sentences， the nearest neighbor strategy combining with Convolution Neural Network （CNN） was used to obtain aspect neighborhood information， so as to reduce the influence of far irrelevant information on the model. At the same time， the grammatical information of sentences was introduced to increase the dependency between words. After fusing the two features， Mask and attention mechanism were used to pay special attention to the aspect information and reduce the interference of useless information to the sentiment analysis model. Besides， in order to evaluate the influence degree of contextual and grammatical information on sentiment polarity， an information evaluation coefficient was proposed. Experiments were carried out on five public datasets， and the results show that compared with the sentiment analysis model AGCN-MAX （Aggregated Graph Convolutional Network-MAX）， the EDNR model has the accuracy and F1 score on dataset 14Lap improved by 2.47 percentage points and 2.83 percentage points respectively. It can be seen that the EDNR model can effectively capture emotional features and improve the classification performance.

The abundant structure and association information contained in Knowledge Graph (KG) can not only alleviate the data sparseness and cold-start in the recommender systems, but also make personalized recommendation more accurately. Therefore, a knowledge graph driven end-to-end recommendation model of graph neural network, named KGLN, was proposed. First, a signal-layer neural network framework was used to fuse the features of individual nodes in the graph, then the aggregation weights of different neighbor entities were changed by adding influence factors. Second, the single-layer was extended to multi-layer by iteration, so that the entities were able to obtain abundant multi-order associated entity information. Finally, the obtained features of entities and users were integrated to generate the prediction score for recommendation. The effects of different aggregation methods and influence factors on the recommendation results were analyzed. Experimental results show that on the datasets MovieLen-1M and Book-Crossing, compared with the benchmark methods such as Factorization Machine Library (LibFM), Deep Factorization Machine (DeepFM), Wide&Deep and RippleNet, KGLN obtains an AUC (Area Under ROC (Receiver Operating Characteristic) curve) improvement of 0.3%-5.9% and 1.1%-8.2%, respectively.

In the detection of components with uneven thickness by X-ray, the problems of low contrast or uneven contrast and low illumination often occur, which make it difficult to observe and analyze some details of components in the images obtained. To solve this problem, an X-ray image enhancement algorithm based on gradient field was proposed. The algorithm takes gradient field enhancement as the core and is divided into two steps. Firstly, an algorithm based on logarithmic transformation was proposed to compress the gray range of an image, remove redundant gray information of the image and improve image contrast. Then, an algorithm based on gradient field was proposed to enhance image details, improve local image contrast and image quality, so that the details of components were able to be clearly displayed on the detection screen. A group of X-ray images of components with uneven thickness were selected for experiments, and the comparisons with algorithms such as Contrast Limited Adaptive Histogram Equalization (CLAHE) and homomorphic filtering were carried out. Experimental results show that the proposed algorithm has more obvious enhancement effect and can better display the detailed information of the components. The quantitative evaluation criteria of calculating average gradient and No-Reference Structural Sharpness (NRSS) texture analysis further demonstrate the effectiveness of this algorithm.

In order to improve the recognition ability for online behavior continuous sequences and enhance the stability of behavior recognition model, a novel online behavior recognition method based on Probabilistic Latent-Dynamic Conditional Random Field (PLDCRF) from surveillance video was proposed. Firstly, the Space-Time Interest Point (STIP) was used to extract behavior features. Then, the PLDCRF model was applied to identify the activity state of indoor human body. The proposed PLDCRF model incorporates the hidden state variables and can construct the substructure of gesture sequences. It can select the dynamic features of gesture and mark the unsegmented sequences directly. At the same time, it can also mark the conversion process between behaviors correctly to improve the effect of behavior recognition greatly. Compared with Hidden Conditional Random Field (HCRF), Latent-Dynamic Conditional Random Field (LDCRF) and Latent-Dynamic Conditional Neural Field (LDCNF), the recognition rate comparison results of 10 different behaviors show that, the proposed PLDCRF model has a stronger recognition ability for continuous behavior sequences and better stability.

In view of the problem that data for Named Data Networking (NDN) cache is replaced efficiently, a new replacement policy that considered popularity and request cost of data was proposed in this paper. It dynamically allocated proportion of popularity factor and request cost factor according to the interval time between the two requests of the same data. Therefore, nodes would cache data with high popularity and request cost. Users could get data from local node when requesting data next time, so it could reduce the response time of data request and reduce link congestion. The simulation results show that the proposed replacement policy can efficiently improve the in-network hit rate, reduce the delay and distance for users to fetch data.

For a better trade-off between fault-tolerance mechanism and fault-tolerance overhead in processor reliability research, a fault detection approach for Multi-Processor System-on-Chip (MPSoC) that placed the calculation task of detecting code on redundancy core was proposed in this paper. The approach achieved MPSoC failure detection by placing the calculation and comparison parts of detecting code on redundancy core. The technique required no additional hardware modification, and shortened the design cycle while reducing performance and memory overheads. The verification experiment was implemented on a MPSoC by fault injection and running multiple benchmark programs. Comparing several previous methods of fault detection in terms of capability, area, memory and performance overhead, the experiment results show that the approach is effective and able to achieve a better trade-off between performance and overhead.

To overcome the influence from both the light change and blurring in actual shooting for the three-dimensional reconstruction based on the stereo vision technique, the new illumination-robust combined invariant moments were put forward. Meanwhile, for the purpose of improving the performance of the image feature matching which solely depended on similarity, the dual constraints of the slope and the distance were involved into the similarity measurement, and then the matching process was carried out with their combined actions. Finally the three-dimensional reconstruction of the whole belly contour was built automatically. The parameters of the belly shape obtained by the proposed method can achieve the same accuracy as the 3D scanner and the measurement error with the actual value was less than 0.5cm. The experimental results show that the hardware of this system is simple, low cost as well as fast and reliable for information collection. The system is suitable for apparel design.

In view of the kinematic model of mobile robot, a tracking controller of global asymptotic stability was proposed. The design of tracking controller was divided into two parts: The first part designed the control law of angular velocity by using global fast terminal sliding mode in order to asymptotically stabilize the tracking error of the heading angle; the second part designed the control law of linear velocity by using the Lyapunov method in order to asymptotically stabilize the tracking error of the planar coordinate. By combining Lyapunov stability theorem and two control laws, the mobile robot can track the desired trajectory in a global asymptotic sense when the angular velocity and the linear velocity satisfy these control laws. The experimental results show that the mobile robot can track desired trajectory effectively. It is helpful for promoting the practical application.