Research On APP For Rapid Detection Of Rice Maturity Based On Computer Vision

The machine vision and image processing technologies were used to detect and research the maturity of paddy rice,which is of great significance to improve the market competitiveness of Chinese rice.If farmers pick rice when it is immature or over mature,it will significantly reduce the yield and quality of rice,and affect the final milled rice rate and cooking quality.At present,the determination of rice maturity in China is mostly based on Farmers experience,or based on a series of physical and chemical properties of rice in the laboratory.The above methods are in the stage of research and search reference.Therefore,this paper attempts to establish a simple,flexible,universal and cost-effective method of rice maturity identification by using low demand hardware(mobile phone lens)and deep learning algorithm.The main contents of this paper are as follows.(1)Experimental determination of the relationship between rice maturity and green rice rate during the whole growth cycleThe green rice rate of rice can reflect the maturity of rice.This article mainly collects rice images of four different varieties as the main information source,studies the influencing factors and change laws between maturity and green rice rate through image information,and constructs the green rice rate as an index model for judging maturity,which provides a basis for the development of APP.(2)Big data analysis of the relationship between RGB information and maturity of rice and research on correction methods based on color cardsFirst,a data set of about 20,000 rice images was established through manual screening of the data set,and then through the analysis of the image information of rice,the different RGB information corresponding to different maturity was calculated by using the way of big data statistics.And the algorithm was assisted to predict the maturity of rice by making the way of "colorimeter",which can make the algorithm have adaptive ability and better robustness.(3)Designed and developed a rice maturity prediction APP based on Python and Java programming languageBased on Pycharm programming software,using Python programming language and Java programming language as tools,the obtained rice image is regularized and normalized,and a series of data preprocessing work such as noise processing and image enhancement are used for the data.Combined with BP neural network for model training,a deep learning model is established.Compared with the way of manually judging the maturity of green rice,the maximum absolute error is 8%,and the maximum relative error is 13%.The recognition rate of rice position has reached 99.1%,and the recognition rate of contrast color plate position has reached 99.3%.Finally,APP is written through Java,and an interface is provided for users to use.