Research And Application Of Remote Sensing Image Machine Learning Platform Based On Container Technology

At present,artificial intelligence is increasingly integrated with all walks of life,and the combination of natural resources and artificial intelligence is in the ascendant.At present,remote sensing image has made great progress.With its advantages of large data,high timeliness,variety of types,rich information and significant utilization value,the related research on big data and machine learning is in full swing.Machine learning platform occupies the central position of the ecological link of artificial intelligence industry,and is the software and hardware foundation of the development of artificial intelligence industry.The status of machine learning platform is becoming more and more prominent,and its function and value are more and more concerned.The emergence of machine learning platform can make the training model iterative quickly and the target data can be produced efficiently.The machine is allowed to undertake most of the work.Only the manual needs to check or modify the classification interpretation results to obtain the data that meets the requirements.The common developers with basic machine learning knowledge background can also skillfully carry out modeling work of common machine learning models according to specific business,so as to improve the work efficiency,reduce the difficulty of machine learning model application and save the development cost of algorithm model.In order to solve the lack of one-stop machine learning business support platform in the field of remote sensing,machine learning is generally faced with;The computing resources are scattered and lack of overall planning and management;The paper combines docker container,kubernetes container arrangement service and other technologies to design and implement the function design,platform management model and platform application model of remote sensing image machine learning platform.A one-stop remote sensing image machine learning platform with high specificity,strong adaptability and excellent usability is built.Based on the existing remote sensing monitoring data of mine,the paper designs the experiment of typical mine feature identification,constructs a deep learning model of typical mine feature recognition based on mask r-cnn,and verifies the feasibility of the platform according to the standardized operation steps of the platform.The test shows that the machine learning platform is superior to the traditional single machine training environment in efficiency,and has incomparable advantages in the use of multi-user,multi task and multi environment.In terms of function,the platform divides the modules of system management,project management,environment management,mirror management,task management and monitoring management,and puts forward clear business process specifications for platform administrators,model developers and business application personnel.Taking the typical mine figure intelligent identification as an example,this paper implements the application example of the whole process of the specific depth model based on the machine learning platform,which strongly supports the practice research of artificial intelligence and machine learning in the remote sensing field.