| Monitoring the species and quantities of pests in orchards is the basis of precise pest control in orchards.Current pest monitoring is mostly completed by plant protection personnel on-site investigation and statistics,which is time-consuming,laborious and limited coverage.With the continuous development of information technology,solar energy,Internet of Things and image recognition technology provide a new development direction for pest monitoring system.Among them,solar energy technology provides electricity to meet the needs of field use of monitoring system.Internet of Things technology provides a high,accurate and fast implementation scheme for information collection and transmission.Image recognition technology completes automatic classification and counting of pests.Based on the analysis of pest monitoring and image pest identification in the Internet of Things at home and abroad,combined with solar charging technology.Based on the analysis of pest monitoring and image pest identification in the Internet of Things at home and abroad,combined with solar charging technology,a solar pest monitoring system for orchards based on Internet of Things and image recognition was developed with peach stem borer,oriental fruit moth,peach fruit borer and fall webworm as pest research objects.The monitoring system is composed of the Internet of Things solar pest monitoring terminal and remote monitoring cloud platform.The monitoring terminal is developed with STM32,4G communication technology and improved MPPT photovoltaic charging control technology.It realizes the functions of high-efficiency photovoltaic charging,environmental information monitoring,pest trapping,pest image acquisition and remote information transmission.The remote monitoring cloud platform is based on Aliyun server,which is responsible for information monitoring,data storage,monitoring terminal control and pest image processing and recognition.Aiming at the problem of low charging efficiency and instability caused by traditional MPPT photovoltaic control algorithm,an adaptive step-size perturbation observation MPPT photovoltaic control algorithm is proposed,which can track the maximum power point faster and more accurately by adaptive step-size adjustment,so as to meet the requirements of efficient and stable photovoltaic charging.Due to the existence of multi-objective insect pests in orchard pest image,a multi-objective insect pest segmentation scheme is formed by analyzing and comparing image gray-scale,image enhancement,background removal,morphological processing and connectivity domain marking.Based on the analysis of the morphology,color and texture characteristics of the target pests,a feature selection method isestablished,which takes the complexity,eccentricity,duty cycle,elongation,sphericity,color moment and gray level co-occurrence matrix of the target image as the characteristics of the pests.Aiming at the slow convergence speed,long training time and low recognition rate of BP neural network,an improved BP neural network algorithm(AGA-IBP)is proposed.By introducing momentum term to improve BP neural network itself and using adaptive genetic algorithm to optimize the initial weights and thresholds of BP neural network,the convergence speed of model network is accelerated,the possibility of local minimum is reduced,and the accuracy and efficiency of pest identification in orchards are improved.Finally,an experimental test was carried out on the monitoring cloud platform.The results show that the output power of photovoltaic cells is increased by 5.1% compared with the traditional disturbance observation method.The average error rate of pest counting in orchards is 2.6%.The recognition rate of common pests in orchards is 73%.The recognition rate of pear Oriental borer is 65%.The peach Oriental carnivore is 66%.The recognition rate of Oriental moth and white moth was 77%respectively,which validated the validity and reliability of orchard pest monitoring system. |
PDF Full Text Request