Research On Greenhouse Crop Environmental Situational Context-awareness Based On Deep Learning

With the vigorous development of China’s facility agriculture,the built-up area has been in the forefront of the world,and the proportion of multi-span greenhouses has increased year by year.However,the greenhouse supporting monitoring and decision-making system has a large gap with the developed countries,and a large amount of monitoring data is not efficient in the greenhouse production process.The use of decision-making systems is too dependent on experience and not intuitive enough,resulting in excessive human-computer communication costs,limited intelligence and automation.At this time,the development and popularization of context-aware technology provides a new solution for intelligent decision-making in greenhouse environment.Therefore,this paper proposes a context-aware system based on greenhouse environment for intelligent decision-making in greenhouse production,which can automatically perceive greenhouse crop environmental changes and give real-time decision information and regulatory recommendations.The existing situational awareness framework and system development are mainly concentrated in the military or smart office.The use environment and greenhouse production are very different.Therefore,after reading a large number of context-aware system related literatures,this paper applies the current status and advantages and disadvantages of each sensing system.The analysis,the existing situational awareness framework and middleware(perceived framework decision module)to improve and innovate,remove redundant modules such as semantic analysis,increase the actual production demand module such as greenhouse decision information database,and the perception system framework inherits the layered design concept.It can respond flexibly to different regulatory requirements and increase system generalization capabilities.The system innovatively uses the deep learning model as an interpreter for middleware,which is responsible for data fusion of multi-source heterogeneous perceptual information,effectively making up for the inaccuracies of traditional fusion models with low precision and single type of converged data.The greenhouse environment data fusion model of the deep classification network and the net photosynthetic rate prediction model of the greenhouse crop based on the deep regression network.Through performance verification,the deep classification environment fusion model has a sensitivity of 98.00% for the environmental status of greenhouse crops,and the correlation between the predicted value of the net photosynthetic rate of the greenhouse crops and the true value of the deep regression prediction model is as high as 0.991.Further,combined with other evaluation indicators,the deep learning model can be indicated.Combined with context-aware systems is feasible and effective.Finally,the Tkinter GUI toolkit built in Python was used to develop the greenhouse context awareness system client,which centrally displays and manages the environment awareness function and the photosynthetic prediction function.The display content includes the greenhouse area perception result,the greenhouse overall perception result and the greenhouse regulation decision information.The client functions include the sensing data storage function,the sensing system personalization setting function and the data prediction function,and the login interface and the registration interface,and the interfaces can be reasonably jumped to form a complete greenhouse situation sensing system.Through the verification of the system function,each module has complete functions,and the constructed context-aware framework and middleware can sense the state of the greenhouse environment in real time and meet the expected requirements,indicating that the work done in this paper is feasible and effective.