| With the development of modern agriculture,the management mode of apple orchards tends to be modernized,and the spraying operation of water,fertilizer,and medicine by the new canopy spraying method is also gradually developed.However,on the whole,there are some problems such as low automation and intelligence degree of spraying system,small scale of operation,and lack of effective integration with environmental information for spraying control.In this research,the intelligent canopy spraying system of fusioning microdomain environment information in apple orchards as the research content,the canopy spraying automatic control technology for relieving harm and reducing pesticide,the content and fusion method combined with the environmental information,and the corresponding intelligent algorithm and control method are analyzed theoretically,the model algorithm is also constructed,and the experimental research and verification are carried out.The main research contents are the acquisition and intelligent analysis method of multi-source environmental information,the integration of environmental information and feature extraction algorithm for spraying decision control,and the automation and intelligence of canopy spraying system,so as to provide technical reference for the intelligent canopy spraying system integrated with environmental information.Starting from the requirement of timely,fast and accurate spraying operation of the canopy spraying system,this study analyzes the basic composition,functional characteristics,and working conditions of the canopy spraying system.The software and hardware methods required for automation control,such as related technology,software control algorithm logic,process design,etc.,have been studied in depth.On the basis of realizing the automation of the spraying system,a canopy spraying system fused with environmental information is proposed to meet the timing needs of timely spraying operations and the algorithm to realize its intelligent control.The main research work carried out is summarized as follows:(1)The automatic control method of canopy spraying system is designed for the problems of low automation and intelligence of canopy spraying system and small scale of operation.The automatic functions such as the design of spraying control process,the selection of master and controlled devices required for automatic control of valves,the control of interaction process of wireless communication,the coding and decoding format of communication protocols,the selfchecking of data in reliable communication,the automatic request retransmission mechanism,the linkage control of multiple groups of valves,the control of operational parameters,the automatic generation and execution of commands from the master,etc.are completed,and the intelligent algorithm and control logic are further constructed.By dividing multiple plots to design communication protocols,we realize the management of multiple plots at one controlled end from the design of pipeline layout and control algorithm,and realize larger scale control from the control logic.By optimizing the structure of the system to improve the utilization rate of the pipeline,it realizes intensive management control.The experimental results show that the algorithm and automated intelligent solution proposed in this paper can meet the needs of precise spraying operation of canopy spraying system,and the scheme is reasonable and feasible.(2)Aiming at the problems of low time resolution and quantitative accuracy,unreliable selection,and uncertainty in predicting the appropriate environmental temperature,humidity,and wind speed required for fast and high-quality spraying operations of canopy spraying systems,this research proposes a multivariate joint prediction,multi-step time step,and short time interval prediction model for micro-domain environmental temperature,humidity,and wind speed in apple orchards.The method first constructs a convolutional network to obtain the changing trends and relationships among multiple environmental factors from historical data,and based on this,a joint prediction model of temperature,humidity,and wind speed with multi-step time steps,multiple variables,and short time intervals every 15 minutes is constructed by introducing forward validation,long and short-term memory networks,and automatic coding methods to convey the time-series information,and effectively reduced the multi-step accumulated errors in multi-step prediction.The test results show that compared with the comparative model,the constructed model achieves the best prediction accuracy,with RMSE = 1.88 in temperature prediction,RMSE = 3.85 in humidity prediction,and RMSE = 0.95 in wind speed prediction,it outperforms other comparative models in the stability of continuous prediction.The model provides an accurate reference model for the joint prediction of temperature,humidity,and wind speed in the apple orchard micro-domain at short intervals and also provides data support for the environmental prediction information required in intelligent spraying decision-making.(3)Aiming at high temperature cooling spraying operation on the apple fruit surface,the problem of insufficient real-time acquisition and accurate measurement of apple surface temperature on low-computing embedded devices in orchards,the machine vision and unsupervised image target automatic classification and extraction was proposed.The combined intelligent solution makes up for the insufficiency of the real-time measurement method of the surface temperature of the apple.The technique integrates micro-embedded devices,the thermal/RGB imager,through the combination of machine vision and unsupervised learning technology,it realizes automatic acquisition of images and temperatures,automatic recognition of apple images in the image target area,automatic temperature correction and extraction function,etc.It has been verified by experiments that the algorithm achieves the lowest omission rate and segmentation error rate in target apple recognition compared with other comparison algorithms,with an average of 12.09 % and 0.13 %,respectively.When the target emissivity is 0.95,the best temperature measurement value can be obtained,and the RMSE measured at a distance of 1 m is1.29 ℃.At the same time,the test also concluded that there is a maximum temperature difference of 18 ℃ between the surface temperature of the apple and the ambient temperature.The implementation of this algorithm provides a solution for the real-time measurement of apple surface temperature.At the same time,the performance of this method also offers accurate apple surface temperature data for the intelligent spraying decision of the canopy spraying system integrating environmental information.(4)Aiming at the problem that the multi-constraint conditions of the spraying operation with environmental information fusion are relatively complicated to select by manual analysis.There is a lack of available models to analyze and make spray classification decisions automatically.An integrated learning intelligent classification algorithm model fused with environmental information is proposed to be based on the characteristics of environmental information and automatically make a classification selection method for spraying.First,the obtained data from multiple data sources is used as a data set.According to the limitation of suitable spraying conditions,a feature data set is established and the spraying classification information is labeled.Then the classification decision result is obtained through the constructed ensemble learning optimization model that integrates the environmental feature information.The model’s performance is compared and verified using the ensemble learning model of environmental feature fusion and other comparable models that were constructed.The proposed model comprehensively surpasses the comparison model in indicators such as recall rate,precision rate,F value,MCC,OA,and IOA.The test results verify the validity of the intelligent algorithm model and provide an effective method for selecting spray classification with multiple constraints of the canopy spray system integrating environmental information. |
PDF Full Text Request