Offensive speech has a serious negative impact on social stability. Currently， automatic detection of offensive speech focuses on a few high?resource languages， and the lack of sufficient offensive speech tagged corpus for low?resource languages makes it difficult to detect offensive speech in low?resource languages. In order to solve the above problem， a cross?language unsupervised offensiveness transfer detection method was proposed. Firstly， an original model was obtained by using the multilingual BERT （multilingual Bidirectional Encoder Representation from Transformers， mBERT） model to learn the offensive features on the high?resource English dataset. Then， by analyzing the language similarity between English and Danish， Arabic， Turkish， Greek， the obtained original model was transferred to the above four low?resource languages to achieve automatic detection of offensive speech on low?resource languages. Experimental results show that compared with the four methods of BERT， Linear Regression （LR）， Support Vector Machine （SVM） and Multi?Layer Perceptron （MLP）， the proposed method increases both the accuracy and F1 score of detecting offensive speech of languages such as Danish， Arabic， Turkish， and Greek by nearly 2 percentage points， which are close to those of the current supervised detection， showing that the combination of cross?language model transfer learning and transfer detection can achieve unsupervised offensiveness detection of low?resource languages.

Concerning the fact that there are few researches on continuous action recognition in the field of action recognition and single algorithms have poor effect on continuous action recognition, a segmentation and recognition method of continuous actions was proposed based on single motion modeling by combining sliding window method and dynamic programming method. Firstly, the single action model was constructed based on the Deep Belief Network and Hidden Markov Model (DBN-HMM). Secondly, the logarithmic likelihood value of the trained action model and the sliding window method were used to estimate the score of the continous action, detecting the initial segmentation points. Thirdly, the dynamic programming method was used to optimize the location of the segmentation points and identify the single action. Finally, the testing experiments of continuous action segmentation and recognition were conducted with an open action database MSR Action3D. The experimental results show that the dynamic programming based on sliding window can optimize the selection of segmentation points to improve the recognition accuracy, which can be used to recognize continuous action.

For the problem that task scheduling program in cloud computing environments usually takes high response time and communication costs, a Honey Bee Behavior inspired Load Balancing (HBB-LB) algorithm was proposed. Firstly, the load was balanced across Virtual Machines (VMs) for maximizing the throughput. Then the priorities of tasks on the machines were balanced. Finally, HBB-LB algorithm was used to improve the overall throughput of processing, and priority based balancing focused on reducing the wait time of tasks on a queue of the VM. The experiments were carried out in cloud computing environments simulated by CloudSim. The experiment results showed that HBB-LB algorithm respectively reduced average response time by 5%, 13%, 17%, 67% and 37% compared with Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), Dynamic Load Balancing (DLB), First In First Out (FIFO) and Weighted Round Robin (WRR) algorithms, and reduced maximum completion time by 20%, 23%, 18%, 55% and 46%. The result indicates that HBB-LB algorithm is suitable for cloud computing system and helpful to balancing non-preemptive independent tasks.

When a link weight changes in network with Internet Protocol (IP), routing loops may occur. Such loops increase the network latency and cause packet losses, which cannot meet the needs of high-level real-time service. A fast routing micro-loop avoidance algorithm using a weight sequence was proposed. The link weights were reallocated according to the weight sequence so that no loops would occur during convergence phase. In order to calculate the weight sequence, a safety weight interval was defined to describe the condition for avoiding loops, then the safety interval was used to search a set of safety weight ranges. During calculation, the prunning technology was used to reduce search range and improve efficiency. At last, the final weight sequence was obtained from these ranges. The simulation test results using typical network topology algorithm show that in average five times of link weight reallocation can successfully avoid loops in 87% of topologies. In addition, compared with other existing algorithms using iterative adjustment link weights to solve the routing micro-loop, the computational complexity of the proposed algorithm was greatly reduced by an order of magnitude and the computational efficiency was improved by 30%-80%. The proposed algorithm can greatly shorten the calculation time and more efficiently solve the problem of routing micro-loop, which will avoid network latency and packet loss to provide a high level of service quality.

In this paper, a new monocular vision system was proposed to improve obstacle detection capability of indoor mobile robot. In this system, firstly, the Hue, Saturation, Intensity (HSI) color space conversion of images was performed. Secondly, a small target threshold selection method was proposed to segment the images, which enhanced the precision of the image segmentation. Thirdly, the target scene matching method and target projection matching method were used to calculate the change of the target pixel and projection so as to judge whether the target is obstacles or ground graphs. The experimental results show that the monocular vision system is effective and feasible, and this system can be applied to the navigation for small indoor mobile robots.

Against shortcomings and limitations of traditional focused crawling methods, a wikipedia-based focused crawling with page segmentation was proposed. It set up topic vector by category tree and topic descriptive document of wikipedia, which described topic; introduced page segmentation after downloading a web page, to filter noise nodes; took block relevance into consideration when computing the priority of candidate links，making up for limited information of anchor text; and validated whether different detailed degree of topic description would effect the performance of focused crawling or not, via changing the size of topic vector space. Experimental results show that this method is effective and scalable, and within a limited degree, the more detailed the topic description, the more related to the topic the collected web pages are.

The basic theories and main application fields of video watermarking were described,then the fundamental principles and specific challenges of video watermarking were analyzed,and its models and classification of algorithms were expatiated on. Finally the possible directions for further research were pointed out.