Video surveillance system is the integration of computers, networks, communications, and video CODEC, etc. Because of its distributed architecture, parallel image processing and ease of installation and expansion, it is widely used in many fields such as education, transportation and industry. However, there are some challenges of video surveillance applications in smart cities such as large scale of video events, low quality and big delay of video data transmission, and the loss of video surveillance data integrity. In order to solve the above problems, this paper designs a series of optimization algorithms and scheduling strategies based on Unmanned Aerial Vehicle (UAV) cluster. Firstly, we construct a full device coverage network with UAV cluster in heterogeneous communication environment of smart cities. Secondly, we formulate the scheduling problem of UAV cluster as bi-objective fragile bin packing problem, and design an optimal scheduling algorithm with constant approximation performance ratio. The simulation experimental results fully demonstrate the effectiveness, feasibility and robustness of the proposed solution in terms of system life cycle, video decodable frame rate, the ratio of UAV flight time to system life cycle, throughput and delay. INDEX TERMS Smart city, video surveillance, unmanned aerial vehicle (UAV) cluster, scheduler, bin packing, heterogeneous communication.

Abstract-As an open-source distributed programming framework, Hadoop has gradually become popular in industry recently. Its distributed file system (HDFS) enables storing large data with advantages of high fault tolerance and throughput. However, the fact that the current HDFS does not support intra-cloud data encryption yet makes data privacy becomes a key security issue. This paper presents ahybrid encryption method based on HDFS. We adopt symmetric encryption to encrypt and decrypt file blocks at datanodes and use asymmetric encryption scheme to protect the symmetric keys. By this method, we can prevent datanode intruders from stealing user data, while ensuring that clients are lightweight. The experiments show that with and without block encryption algorithm, our solution brings43% and 2% performance degradation compared to the generic HDFS.

No abstract

No abstract

No abstract

No abstract

Due to the low segmentation accuracy and sensitivity to initial contour in image segmentation of CV model, an image segmentation algorithm based on CV model combined with spatial fuzzy c-means was proposed for MRI and CT image segmentation with unclear boundary, artifact and high noise. Based on the rough segmentation of the image by using the fuzzy c-means clustering algorithm in the spatial domain, the initial contour is set by using the clustering information to assist the CV model, and the target region is segmented by iterative evolution. The experimental results showed that when the number of iterations was only 50, the Dice coefficient of our algorithm for segmentation of brain MRI images was 89.17%, 38.9% higher than the traditional CV model. It can be seen that the algorithm has higher discrimination and better segmentation effect for medical images.

No abstract