Design Of Display Platform For High-concurrency Multi-dimensional And Multi-source Unmanned Aerial Vehicle Remote Sensing Data

In the era of big data.unmanned aerial vehicle(UAV)with significant advantages in low-altitude fields has gradually become one of the important tools for acquiring accurate remote sensing data.The data volume of UAV remote sensing data is growing rapidly in the form ofgeometric magnitude.Due to the large amount of data,for remote sensing data from different sources and different time,there is a lack of standardized data processing procedures and reasonable UAV remote sensing data storage and visualization model.In order to better organize and utilize the massive UAV data and take advantage of the UAV remote sensing system,this thesis designs a display platform supporting high-consensus for massive multi-dimensional and multi-source UAV data.Combing distributed clustering with NoSQL technology,this thesis elaborates the system implementation from two main aspects:data storage and caching strategy.With the purpose of vegetation monitoring,reasonable applications are designed.The main work done is as follows:(1)To solve the storage problem of massive UAV data,a data storage structure based on HBase is designed.On the basis of completing the preprocessing of the UAV data,in order to shorten the data I/O time,the Hilbert coding is used to establish the index.According to the multi-dimensional and multi-source characteristics of UAV data,HBase storage table is designed based on data locality.It can store a variety of heterogeneous remote sensing images without modifying the storage model.In addition,low-configuration and multi-server clusters are used to storage data,which can meet the growing demand for storage space by increasing the number of servers.(2)To speed up the response speed of map browsing and improve the concurrency of the system,a Redis-based caching strategy is designed.The memory-based Redis is used to implement the caching mechanism,which can reduce direct access to the hard disk database and improve response speed.Based on the analysis of user behavior and tile pyramid model,a cache strategy of neighborhood prefetching is proposed.In terms of breadth and depth.the hotspot data adjacent to the display area is cached into Redis in advance,and the allkeys-lru algorithm is set as the replacement strategy.Experiments show that the neighborhood prefetching cache strategy can significantly improve the cache hit ratio compared to the traditional single tile as the cache granularity.(3)To complete the platform development more conveniently,the system adopts the Model-View-Control(MVC)mode as the development framework.The server side uses an event-driven asynchronous I/O model to improve system concurrency.OpenLayers,an open source JavaScript library,is used to implement application development in client.In addition to the basic display of visible light image,the platform also supports serialization display of time-dimensional image and dynamic combination of multi-spectral data in different bands.In addition,the platform also provides basic measurement.statistics,generation and uploading shape file for agricultural analysis.(4)To test the concurrency of the system.test environment is set up and multiple comparative experiments are designed.The system concurrency performance is evaluated from three aspects:cache hit rate,average response time and server throughput.The experimental results show that the proposed solution has higher concurrency performance compared with the traditional method.In addition,the system has the advantage of supporting dynamic expansion of data.