Design And Implementation Of Orchard Smart Irrigation System Based On Phenological Period Recognition

Most orchards in China are located on hillsides,hills and other places,with insufficient water sources,difficult water supply and poor soil.Traditional irrigation and fertilization methods are often used,resulting in problems such as low irrigation water utilization rate and abuse of chemical fertilizers;the water and fertilizer requirements of fruit trees vary significantly at different growth and development stages.Phenological period is an important factor affecting water and fertilizer demand.In order to realize the timely and appropriate irrigation and fertilization of fruit trees in different phenological stages,this paper constructs an orchard intelligent irrigation system based on phenological stage recognition.First,in order to realize fast and effective recognition of fruit tree phenology,a lightweight phenological recognition method based on improved YOLOv5 proposed using the apple phenological data set as a sample.The peach tree phenological stage data set collected through the network is used as the test sample,and the above-mentioned apple phenological stage recognition model is used as the starting point,and the transfer learning training strategy is used to realize the effective recognition of the phenological stage of a small sample of fruit trees.The method of phenological recognition integrated with the irrigation system,and an intelligent irrigation system for orchards based on phenological recognition constructed.The system makes decisions on water and fertilizer based on the results of apple phenological recognition,corresponding planting strategies and orchard environmental conditions,and realizes automatic irrigation and precise fertilization.The main research contents of this paper are as follows:（1）Apple lightweight phenological recognition method based on YOLOv5Aiming at the practical challenges of complex orchard environment,small detection target and limited performance of edge computing platform,a lightweight apple phenological recognition method based on the improved YOLOv5 model proposed.Firstly,based on the improved ShuffleNet V2,a feature extraction network is built to reduce the model size,combined with the collaborative attention mechanism to improve the feature extraction ability,and a small target detection layer is added to improve the small target detection ability.The experimental results show that the improved model only loses 0.7% mAP,the computational complexity is only 16.5% of the original,and the single recognition time under the CPU platform is reduced by 57.2%.The average accuracy rate of phenological recognition is 97.4%,which can realize fast and effective recognition of apple phenological period.（2）Fruit tree phenological recognition method based on transfer learningThere are various types and varieties of fruit trees planted in China,and there are similarities in appearance and texture characteristics.In view of the practical problems of long collection period and small data volume of phenological data sets.A small sample fruit tree phenological recognition method based on transfer learning is proposed.The phenological data set of peach trees collected through the network is used as the test sample,and the abovementioned apple phenological recognition model is used as the starting point,and the transfer learning training strategy is used to realize the effective recognition of the phenological period of a small sample of fruit trees.The test results show that the average accuracy of the transfer learning model is improved by 5.7% compared with the original model,and the model convergence speed is significantly improved.（3）Design and realization of smart irrigation system in orchardIn order to realize automatic irrigation and precise fertilization,an orchard intelligent irrigation system is designed.The system consists of four parts: acquisition module,edge computing module,server module and irrigation execution module.The system makes water and fertilizer decisions and realizes automatic irrigation according to the recognition results of fruit tree phenology,corresponding planting strategies and orchard environmental conditions.The orchard intelligent irrigation system can automatically irrigate and accurately fertilize apples at different growth stages on demand.The test results show that the orchard intelligent irrigation system can automatically irrigate based on decision-making information and soil moisture.