Research And Realization Of Integrated Control System Of Water And Fertilizer In Greenhouse Based On Speech Recognition Technology

The yield and quality of crops in greenhouses are closely related to the water and nutrients absorbed by crops during their growth period.Integrated irrigation with water and fertilizer is currently a key technology for achieving efficient planting in greenhouses.However,existing systems still have problems with cumbersome operation and low intelligence,which are not suitable for the current situation and needs of greenhouse planting in China,Ordinary farmers are prone to misoperation and excessive irrigation of water and fertilizer.Therefore,in order to improve the efficiency of water and fertilizer utilization and reduce environmental pollution,this article designs and develops an integrated intelligent irrigation system for greenhouse water and fertilizer based on speech recognition technology,achieving precise fertilization in greenhouse,laying the foundation for promoting the standardization,intelligence,and large-scale production of smart agriculture.The research content and experimental results are as follows:(1)Researching existing speech recognition technology network models,integrating Dropout technology,CTC algorithm,and CNN to theoretically demonstrate and derive the network model structure,studying the construction method of speech recognition models based on DenseNet,and designing model network structure parameters;By collecting and producing audio samples of commonly used commands for the integrated water and fertilizer system in production practice,a dataset of voice and instruction corresponding to water and fertilizer irrigation scenarios was obtained;Based on the DenseNet speech recognition model and network structure parameters,model training was completed on the samples and a converged and portable network model file was obtained.The network model was evaluated and validated by evaluating the samples.The experimental results showed that the error recognition rate of 1400 samples was about 2.77%,and for speech with too long input,the error rate gradually increased.After using regular search,it can meet production needs.(2)Based on the speech recognition hardware foundation of the development board core of Nvidia JetsonTX2,the node speech recognition unit is designed,and the convergence model is transplanted to the development board to complete the construction of intelligent speech recognition module;A water fertilizer integrated intelligent voice irrigation system has been developed with the STM32F429IGT6 main controller as the core,touch screen,relay,and various sensors as the hardware foundation;Finally,the intelligent speech recognition module,integrated water and fertilizer control system,and cloud server were combined to achieve the function of controlling the integrated water and fertilizer system through both remote user input voice and on-site user input voice.(3)Based on the integrated intelligent voice irrigation system of water and fertilizer,experimental testing and conclusion analysis were conducted on the system’s function and performance.The test results showed that the error rate of the system’s voice recognition function was basically consistent with the model validation analysis results,and the error rate of instructions was relatively low;Secondly,in terms of remote control delay performance,the comparison of control system response delay between Ethernet and wireless Wi Fi was emphasized.The experimental results showed that the Ethernet delay was less than 1.5 seconds,and the Wi Fi delay was less than 2 seconds.Overall,Ethernet had higher stability and lower delay compared to wireless Wi Fi.When both were connected simultaneously,the instability and delay increased with the increase of background traffic.In terms of on-site user delay,the experimental results show that the total delay of the control system response decreases significantly,but the delay of speech recognition increases,with the overall delay less than 1 second.The methods to improve the delay are focused on improving the calculation value of the node development board and optimizing the network model.