Distributed Storage And Retrieval Of Multi-layer Vision Data Of Tobacco Fields

In accordance with problems that the present remote image acquisition of crops is lowtimeliness, the weak ability to monitoring and poor acquisition efficiency, and the usage of thetraditional distributed database to store and manage massive agriculture data is poor efficiency,and has weak ability of storage and higher pressure to hardware, our research aims at realizingthe distributed storage and retrieval of multi-layer vision data of tobacco fields, and webcamsare used as the image acquisition equipment to capture the tobacco data. Through the study ofstorage technology of massive data and image retrieval technology, a framework combiningthe hadoop distributed file system and relational database is proposed. This framework workswith text-based image retrieval and perceptual-hash-based image retrieval neatly, and itrealizes distributed system management of multi-layer vision data of tobacco fields.The mainly contents and achievements of the research are listed below:(1) A method based on the framework of the hadoop distributed file system is proposedand it can realize the storage of massive data. Also combining the relational databasetechnology, this method realized efficient image storage and solved the problems of the weakability to store massive data, higher pressure to hardware and high costs.(2) Combining text-based image retrieval and perceptual-hash-based image retrieval, themethod realized the image retrieval of multi-layer vision of tobacco fields, and solved theproblems that text descriptions cannot fully express the image content and cannot realizeimage similarity retrieval based on image visual features.(3) The framework employed Java language and Structs2to design and realize thedistributed management of multi-layer vision data of tobacco fields. The captured multi-layervision data are distributed stored and managed through our framework. Experimental resultindicated that the system could provide effective storage and retrieval for images and videoscaptured by22webcams in the experimental station of Northwest A&F University.Our research can provide efficient technical method and practical support tool foragricultural remote prompt analysis, agricultural production decision, crop yield assessmentand effective control method on disease and insect pests.