| Video as an important source of modern data,the importance of related technology has become increasingly prominent.Although the current video technology in agriculture,industry,military,security,transportation and other scenes have been more common application,there are many problems in the Internet of video technology to be resolved,such as unified interface,weak front-end computing power,not targeted data for users.In most cases,all video streams go through the sensing layer,the transport layer to the processing layer and then the application layer,usually accompanied by a certain delay,affecting the real-time video surveillance and Internet processing efficiency.In order to facilitate the forwarding of video,reduce the delay of video and image data processing,improve the efficiency of data transmission and data processing in Internet of Things,this paper proposes an embedded video middleware technology based on process in field.Through low-power consumption and low-cost,high-performance embedded heterogeneous multi-core computing system for the on-site video capture,processing and forwarding.The main works are listed as follow:(1)We study on the current research status of video surveillance technology,and analyze the existing problems.An embedded video middleware solution is proposed based on Epiphany low power heterogeneous multi-core architecture in this paper.(2)We study and execute the basic algorithm(matrix operation and DCT algorithm)which is widely used in image and video computing on heterogeneous multi-core embedded platform.The some main parallel programming models are analyzed: the programmability and performance of Epiphany SDK,MPI,OpenMP,OpenCL in Epiphany parallel heterogeneous multi-core architecture.We analyze the important influence of inter-core communication on the application performance of Epiphany architecture in this paper.(3)Based on Epiphany architecture,a VLC+DPM video forwarding and pattern recognition embedded middleware is proposed.On the basis of studying the basic algorithm of image video computation,the DPM algorithm is analyzed and transplanted into Epiphany architecture.Through high-performance low-power embedded multi-core computing architecture,real-time front-end site to deal with the completion of videoprocessing and transmission.The experimental results show that the performance of 512 * 512 matrix multiplication in embedded middleware system achieves the processing level of the Core i7 4790 K computer,which is a low-power consumption,low-cost video real-time middleware.Meanwhile,our method provides a useful reference and reference for reducing the data processing delay and improving the efficiency of data processing. |
PDF Full Text Request