Research On Fine-grained Optimization Of Sensor Coverage Based On Deep Neural Network

Small hydropower has a measurable impact on rural social and economic development, energy conservation and disaster relief emergency. With the application of remote monitoring system, hydropower gradually realized the target of "unmanned on duty". Nowadays, accidents still occurs frequently due to the lack of safety regulation in High-risk area, from which higher requirements for the coverage quality should be proposed.In the case of the existing research, the majority of target coverage problems are concentrated on static targets. Thus, the algorithms on fine-grained coverage of wireless video sensor network have a significant sense of improving sensor’s monitoring performance as well as ensuring the safety of small hydropower.The main contributions of this thesis are as follows:1.This paper analyzed behavior data of moving target and defined a series of single risk index. AHP is used as a guide and the interaction relationship between indicators are taken into consideration. At the same time, the paper used a neural network instead of expert decision, building an integrated risk model.2.For the problems of information duplication in collected data of moving targets, this paper put forward a set of solutions. It used a distributed system to carry on de-duplication and dimension reduction in timing data of path, which can reduce the energy consumption during transmission process.3.Considering different covering requirements of different targets in fine-grained covering, this paper designed a multiple targets dynamic probability covering algorithm, MDKCA based on probability forecasting model. Though referencing the algorithm of multiple moving targets with K coverage and optimizing the original computing model, the performance of the algorithm was improved.4. This paper implemented a wireless sensor network monitoring system in small hydropower based on C/ S. And this system has a series of functions contains the factory management, data analysis and decision management.This paper extends the research on moving target with K level coverage and proposes different requirements of coverage quality for different moving targets. In order to guide sensor network coverage, this paper evaluates the level of risk of each moving target through building an integrated risk evaluation model. In view of the existing problems, this paper elaborates on a complete process, including the problem definition, mathematical model, algorithm and simulation experiment. Finally a wireless video sensor monitoring system was developed to detect the accuracy of the algorithm.