Research On Multi-node Cooperative Vision Sensor Network Architecture And Power Allocation

Wireless visual sensor network(WVSN)is an extension of traditional wireless sensor networks,which consists not only of scalar sensors such as temperature and humidity sensor,but also of visual sensor such as cameras.Compared with traditional wireless sensor networks,WVSN extends monitoring information to visual level and enriches the content of monitoring information.Also,cooperative communication technology,as an extension of multi-input and multi-output technology,has become a research hotspot in recent years.Then,by introducing cooperative communication technology into wireless vision sensor networks would form a virtual multi-antenna environment,which can bring diversity gain,extend coverage and then improves network's performance.Meanwhile,this dissertation also studies the power allocation technology based on deep neural work in emphasis,which is more suitable for above cooperative WVSN.The major works are outlined as follows,(1)By designing three-level cooperation between ordinary nodes and camera nodes,a multi-node cooperative heterogeneous wireless vision sensor network architecture is proposed.So,the energy consumption pressure of image acquisition and transmission of camera nodes is decomposed into ordinary nodes,which can reduce the energy consumption and then prolong the life cycle of the whole wireless vision sensor network.(2)Based on above multi-node cooperative heterogeneous wireless vision sensor network architecture,this dissertation also focuses on the cooperative transmission of images,and gives the performance comparison of peak signal-to-noise ratio and mean square error of three communication modes,namely direct transmission,amplificationand-forwarding cooperative communication and decoding-and-forwarding cooperative communication under two power allocation scheme respectively.(3)An optimal power allocation scheme based on deep neural network(PADNN)is proposed,in which the fading factors about source,relay and destination nodes are used as input train data,ergodic capacity of above WVSN is used as optimal goal.The simulation results show that the well-trained deep neural network can allocate power between source and relay nodes very quickly,and then this PWDNN algorithm outperforms equal power allocation one in ergodic capacity in different relay locations,which is more suitable for cooperative WVSN.