Research On Field Rice Yield Estimation Method Based On Deep Neural Network For Rice Ear Recognition

Rice ear is an important part of rice,so it is very important to study and analyze rice ear in field.Rapid detection and counting of rice panicle is of great significance for monitoring crop growth,estimating yield and analyzing plant phenotypic characteristics.Accurate identification of rice ears in paddy fields is one of the important steps to estimate the overall rice yield in paddy fields.At present,the traditional counting of rice ears in paddy fields is mainly manual counting.It takes a lot of time and labor cost to collect the data of rice ears in fixed area in the field,and most of the rice are very similar in appearance,which is difficult to distinguish.The subjectivity of agricultural workers has higher requirements.It is of practical application value to count the number of panicle and estimate the yield of paddy rice by means of agricultural informatization.In this paper,deep neural network is used to count rice in the field,and it is actually applied in agricultural production to predict rice yield,which improves the efficiency of rice-related work of agricultural workers in the field to a certain extent.The specific contents are as follows:(1)Taking rice in the field as the research object,data collection of rice panicle and rice image was carried out in three ways,namely,graying,image flipping and contrast increasing.Label Img software was used to mark the target of rice panicle in the rice panicle image after data enhancement,so as to construct the rice panicle image data set.(2)The YOLO v5 deep neural network model was used to identify and count rice ears in paddy fields,and the SE channel attention mechanism was added into the network structure to optimize the model,so as to improve the detection accuracy of the model.By comparing the collected rice ear data set with the original YOLO v5 network,SSD network and YOLO v3 network,the optimized YOLO v5 network model has an accurate detection rate of 88.6% for rice ears,which is better than other comparison networks and can meet the requirements for the accuracy of rice ear identification in paddy rice yield estimation.(3)The quantity and weight of rice per unit area of paddy field were measured manually,and the relationship between the quantity and weight of rice per unit area of paddy field was determined through the regression analysis of the quantity and weight of rice per unit area,and the rice yield estimation model of paddy field was reversed.Java language was used to develop mobile client system for paddy rice yield estimation,and My SQL associated database management system was used to realize mobile paddy rice count and yield estimation,which was convenient for paddy rice production and experimental activities.