The Application And Research Of Crop Growth Monitoring Based On WSN And Monocular Vision

Up to now,agriculture is becoming more accurate and more informationalized.It is of great significance to apply wireless sensor network(WSN),machine vision,image processing and data mining technology to monitor crop growth and to study the correlation between environment and crop growth.Firstly,this thesis introduces the current research situation and significance of crop growth monitoring which adopts four kinds of technologies mentioned above.Secondly,this thesis introduces an environmental monitoring system of crop growth based on WSN-and its deployments in fields,thereafter seven kinds of environment indicators(illumination intensity,soil temperature,soil humidity,air temperature,air moisture,air CO2 concentration and soil alkalinity)and their correlation are analyzed.Thirdly,this thesis puts forward one non-contact detection method of individual crop height and width based on monocular vision,furthermore the height,width and their ratio of individual corn,the length,width and their ratio of corn leaf are analyzed and the regression models of these parameters combined with environment indicators are established,respectively.Meanwhile,this thesis also compares two different types of regression models.Then,as one of the crop growth phenotypic parameters,the measurement of covering area is introduced and monocular area detection errors based on the image plane are studied.The experiment results show that different indicator shows individual characteristics and there are some certain correlations among different indicators.Detection error of individual crop height and width based on monocular vision ranges from 0.95%to 4.76%and from 0%to 6.73%,respectively.The detection error of covering area of green crop ranges from 0.76%to 13.42%.Two methods both can obtain the ideal results.The correlation coefficients between two types of crop growth phenotypic parameters and environmental indicators both can reach more than 0.8.The correlation coefficient of three-dimensional curved surface fitting of area based on the image plane can reach 0.9758.Finally,the shortcomings of research work and methods which will be optimized further are summarized.