Research On Classification And Recognition Of Growing Period Of Winter Wheat In Central China Plain Based On Deep Feature Learning And Multi-level R-CNN

The growth period of winter wheat is one of the important indicators of wheat production activities in the Central China Plain.Traditional research on the growth period of winter wheat.On the one hand,when automatically detecting and classifying the growth period of winter wheat,traditional image processing methods are used.There are problems such as long detection time and low classification and recognition accuracy,which cannot meet the requirements of smart agriculture.application.On the other hand,the process of extracting crop image features is more cumbersome and has a high degree of dependence on image features.As a result,traditional image classification and recognition methods cannot be widely promoted and applied.This paper takes images of the seedling stage,heading stage and maturity stage of winter wheat in Henan Province in the Central China Plain as the research object,and uses the idea of combining deep feature learning and image processing technology to construct a multi-level R-CNN winter wheat.Classification and recognition model of growth period.First of all,because the image information of winter wheat actually collected in the field is very large,the traditional convolutional neural network has too few convolutional layers to extract the deep features of winter wheat.Therefore,the winter wheat image is segmented by the depth-based separable convolutional segmentation model,and the effect of the model is analyzed and verified;secondly,the winter wheat growing period classification and recognition model based on Faster R-CNN is used to solve the problems of the original recognition model.By replacing the activation function of the VGG16 network,using the clustering method to reset the size of the anchor point frame,and introducing the penalty function method in the NMS algorithm to improve the original recognition model;using the multi-level R-CNN to construct the first level The model classifies and recognizes the growing season of winter wheat.Finally,through the comparison of the multi-stage R-CNN model before and after the improvement,it is found that the overall recognition rate has increased by 11.33%,which verifies the feasibility and practicability of this research.The purpose of this study is to determine whether agricultural production guidance such as irrigation,fertilization,and weeding is needed based on the results of different periods identified in the actual production process of winter wheat,combined with the characteristics of the three growth periods of winter wheat,so as to ensure the stability of winter wheat production.