Intelligent Analysis Of Rice Grain Characters Based On Red Light Transmission

With the global warming and the reduction of agricultural cultivated land,agricultural production is facing severe challenges.The growth rate of grain output has slowed down significantly,and the problem of food security has become increasingly serious.As a large agricultural country,it is particularly important to ensure crop yield and quality.Rice is one of the three major food crops in China,and its importance is self-evident.In order to improve the yield and quality of rice,the research on rice grains has become essential.The yield of rice is closely related to the grain type and grain weight of rice grain.Therefore,this paper mainly studies the full and shriveled grain of rice grain and the grain type of rice grain,such as grain length,grain width,grain area and grain perimeter.With the gradual application of image processing technology in the field of crop phenotype,the study of crop phenotype based on red light image has developed rapidly.In this paper,the red light is transmitted to the rice grain,the rice grain image is collected by the high beat instrument,the target detection technology is used to quickly and accurately identify the full and shriveled grain of the rice grain,and the rice grain phenotype algorithm is used to calculate the grain type of the rice grain.This paper mainly studies from the following aspects:(1)Design of rice grain image acquisition system based on red light transmission.A set of rice grain image acquisition system is designed based on red light transmission.The design idea is to lay the rice grains on the red light source board,transmit the rice grains through the red light source of the specified wavelength,and then adjust the high-speed camera to specific shooting parameters,so that the images collected can clearly distinguish the fullness and shrinkage of rice grains,and finally use it to collect the red light transmission image.(2)Improvement of full and shriveled grain recognition and depth learning model of rice grain image based on red light transmission.Taking the collected rice grain image as the research object,a target recognition model is created to recognize and distinguish the full and shriveled rice grains.Faster R-CNN model,its improved model and YOLOv5 model were used to extract rice grain characters for comparative experiments.Among them,the accuracy rate P,recall rate R and F1 score of rice grain using Faster R-CNN model are 91.60%,90.80% and 91.15% respectively.The accuracy P,recall R and F1 score of the improved model were 97.80%,97.20% and 99.00% respectively.The accuracy rate P,recall rate R and F1 score of rice grain using yolov5 model were 94.11%,88.14% and 91.02% respectively.It can be seen that the improved model has the highest accuracy rate P,recall rate R and F1 score.(3)Study on calculation and model application of rice seed grain type based on red light transmission.This application is divided into two parts.First,design QT application interface,which will call the trained network model,and then transmit the input red light through the rice grain image for target recognition.The target recognition will detect the position of rice grain and generate the minimum external rectangular target frame surrounding the target,and then calculate the parameters such as grain length,grain width,grain area and grain perimeter of the detected rice grain through the rice seed type algorithm.Finally,the total number of grains,full number of grains,number of shriveled grains,grain length,grain width,grain area,grain perimeter and other parameters of rice grains are displayed on the QT interface for users to view and analyze;Second,use cloud computing to deploy the network model and grain algorithm to Alibaba cloud,and then design the network call web page,so that users can upload the red light transmission rice grain image on the web page.Then,the model and grain algorithm deployed on Alibaba cloud will carry out target detection and grain measurement,and finally return the calculation results to the web page for users to view and analyze.