Research On Disaster Information Collection And Disaster Data Compression Encoding Method For UAV Emergency Rescue

In emergency rescue,obtaining accurate disaster information from the affected area quickly is a prerequisite for ensuring the correctness of emergency command.However,due to issues such as insufficient basic communication infrastructure and unsuitable channel models in emergency scenarios,there is often a low timeliness of information transmission and insufficient channel resources in emergency communication.To address these problems,this article introduces the concept of collaborative sensing and the freshness indicator,and improves the system’s sensing efficiency by optimizing the collection and transmission speed.It also studies the endto-end dynamic data compression scheme in the signal-to-noise ratio changing environment based on the source-channel joint coding model to improve resource allocation efficiency by optimizing channel utilization.Therefore,this article focuses on the scene of unmanned aerial vehicle disaster information collection and disaster data compression and coding for emergency rescue,and studies how to improve the effectiveness of emergency rescue based on the changing channel status and communication resource characteristics of different locations during the process of disaster information collection and transmission in emergency rescue scenarios with varying channel conditions and limited communication resources.The specific research content and main contributions are summarized as follows:Firstly,in the scenario of using unmanned aerial vehicles to assist in disaster information collection,this article studies how to efficiently collect and transmit disaster information in emergency rescue scenarios with high demands for data freshness and unstable channels in mountainous forest areas.A model of the scene is established based on the forest channel model.By using the decision network of deep reinforcement learning to judge the most reasonable return position of unmanned aerial vehicles based on environmental information,and optimizing the collection trajectory,the overall task time is minimized while maximizing the data freshness received to ensure that all disaster information is collected.The experimental results show that the unmanned aerial vehicle collaborative sensing path planning algorithm proposed in this article based on deep reinforcement learning saves about 65%of time and improves data freshness by about 132%compared with traditional swarm intelligence algorithms,and saves about 57%of time and improves data freshness by about 75%compared with the original deep reinforcement learning algorithm.Secondly,this article further addresses the issue of limited communication resources for unmanned aerial vehicles in emergency transmission scenarios and proposes an adaptive source-channel joint coding model based on deep learning,which is applied to target detection tasks.After the unmanned aerial vehicle collects disaster information,the model dynamically compresses the transmitted information based on the content of the transmitted image and the signal-to-noise ratio of the transmission location to solve the problem of low information transmission bandwidth utilization in emergency scenarios.The experimental results show that when the environmental signal-to-noise ratio is 20dB,using the proposed model for data transmission can save about 50%of the bandwidth resources.In addition,by training an end-to-end data transmission detection model in the target detection task,the target detection model has the ability to adapt to different image features and environmental signalto-noise ratios,greatly improving detection accuracy.Finally,in order to validate the theoretical algorithms proposed in this paper,a smart disaster information collection platform was built based on an intelligent communication and computing service platform for the scenario of emergency rescue and disaster information collection.The server deployed on the cloud is responsible for implementing the algorithm model,and when a request is received from the user side,the server selects the corresponding algorithm based on the specific request and environmental information,and calls the training resources on the cloud for model training or inference.The results are then stored in the database on the server and the unmanned aerial vehicle is assigned with real-time collection tasks.The user can interact with the database to obtain the UAV’s status and service completion status,which are displayed on the UI interface.The experimental results show that this system can effectively reduce task consumption time while ensuring the complete collection of disaster information,thereby improving the efficiency of emergency rescue and having practical application value.