Peach Phenological Stage Recognition Research And System Implementation Based On Ensemble Learning

As China is a major peach growing and consuming country,the intelligent recognition of peach weather stages provides technical support for digital peach orchards and is an important means of freeing up human resources,so the research on peach weather stage recognition methods is of great significance.In this paper,we use convolutional neural networks and ensemble learning methods to construct a peach weather stage recognition model,and select the optimal model by comparison.In this paper,the ensemble model is used as the core to design and implement a peach weather stage recognition system using Django and other technologies.The research content is as follows.(1)Research on peach phenological stage recognition algorithm based on convolutional neural network.In this paper,we construct a peach season recognition model by improving three different convolutional neural networks,namely Alex Net,VGG-16 and Res Net50,with the six peach seasonal stages as the research object.Firstly,we collected the peach weather phase image samples in the field and completed the construction of the peach weather phase dataset through data processing and data augmentation;secondly,we improved the three models by replacing the LRN layer with a BN layer,adding a coordinate attention mechanism and a global average pooling layer instead of a fully connected layer to achieve the purpose of improving the accuracy of the models.The experimental results showed that the recognition accuracy of the three improved convolutional neural networks was 96.88%,97.45% and97.74% respectively,which was higher than the recognition accuracy of the original models.(2)Research on peach phenological stage recognition algorithm based on ensemble learning.An ensemble learning based peach season recognition algorithm is further proposed based on the research of convolutional neural network peach season recognition algorithm.The Alex Net,VGG-16 and Res Net50 models before and after the improvements in Chapter 3were used as two sets of classifiers,and the majority voting method was used as the ensemble strategy.The two sets of classifiers were trained and the results of the classifiers were fed into the ensemble strategy for voting,and the final output was the result of peach phenology recognition.By analysing the experimental results,the ensemble model obtained by using the improved Alex Net,VGG-16 and Res Net50 models as classifiers outperformed the single model in terms of both accuracy and stability.(3)Design and implementation of a peach phenological stage recognition system.In order to achieve intelligent recognition of peach weathering periods,this paper designs and implements a peach weathering period recognition system with an ensemble model as the core,using technologies such as Django and My SQL.The system provides batch recognition of seasonal images,farming advice and technical learning.The system also has the advantages of simple operation,concise interface and easy operation.It is important to promote the development of digital peach orchards.